diff --git a/gen/descriptor.binpb b/gen/descriptor.binpb
index 3fbd733b..e48595ac 100644
Binary files a/gen/descriptor.binpb and b/gen/descriptor.binpb differ
diff --git a/gen/go/fora/v1/fora.pb.go b/gen/go/fora/v1/fora.pb.go
index 591dea53..82cec52f 100644
--- a/gen/go/fora/v1/fora.pb.go
+++ b/gen/go/fora/v1/fora.pb.go
@@ -2794,8 +2794,8 @@ type Offer struct {
// licensing arrangement: each projected LicenseTerm yields its own offer, and
// the Exchange moves that term's price here, so the term inside `terms`
// carries no pricing of its own (the offer.terms.pricing_unset rule). This is
- // the price execute charges, the price a Broker ranks and compares across
- // Exchanges, and the price a metered purchase is charged at. A resource with
+ // the price execute charges, the price an agent compares across Exchanges,
+ // and the price a metered purchase is charged at. A resource with
// multiple alternative terms (e.g. dual-licensed) produces multiple separate
// offers, one per term — never one offer with a "headline" picked among them.
//
@@ -2995,9 +2995,9 @@ type Offer struct {
// rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a
// resource carries many terms, an offer sells one of them.
//
- // The distinction is not cosmetic. `pricing` is what a Broker ranks on and
- // what execute charges, so a term that is not on its own offer has no price
- // a Broker can compare and no offer_id an agent can buy. Fusing several
+ // The distinction is not cosmetic. `pricing` is what an agent sees and what
+ // execute charges, so a term that is not on its own offer has no price an
+ // agent can see and no offer_id an agent can buy. Fusing several
// terms onto one offer makes every term but the first unsellable — a
// dual-licensed resource sells only under whichever arrangement the
// publisher happened to store first.
diff --git a/gen/jsonschema/fora.v1.DiscoveryResponse.schema.json b/gen/jsonschema/fora.v1.DiscoveryResponse.schema.json
index 52242a03..1ee10087 100644
--- a/gen/jsonschema/fora.v1.DiscoveryResponse.schema.json
+++ b/gen/jsonschema/fora.v1.DiscoveryResponse.schema.json
@@ -260,7 +260,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -290,7 +290,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.DiscoveryResponse.schema.strict.json b/gen/jsonschema/fora.v1.DiscoveryResponse.schema.strict.json
index 435f6f48..ee108205 100644
--- a/gen/jsonschema/fora.v1.DiscoveryResponse.schema.strict.json
+++ b/gen/jsonschema/fora.v1.DiscoveryResponse.schema.strict.json
@@ -264,7 +264,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -294,7 +294,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.Offer.schema.json b/gen/jsonschema/fora.v1.Offer.schema.json
index 6d8c4cd8..f83b6d21 100644
--- a/gen/jsonschema/fora.v1.Offer.schema.json
+++ b/gen/jsonschema/fora.v1.Offer.schema.json
@@ -713,7 +713,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -743,7 +743,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.Offer.schema.strict.json b/gen/jsonschema/fora.v1.Offer.schema.strict.json
index 428f9d6c..53f01acf 100644
--- a/gen/jsonschema/fora.v1.Offer.schema.strict.json
+++ b/gen/jsonschema/fora.v1.Offer.schema.strict.json
@@ -725,7 +725,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -755,7 +755,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.OfferGroup.schema.json b/gen/jsonschema/fora.v1.OfferGroup.schema.json
index 8f49eaa1..d70f7baf 100644
--- a/gen/jsonschema/fora.v1.OfferGroup.schema.json
+++ b/gen/jsonschema/fora.v1.OfferGroup.schema.json
@@ -260,7 +260,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -290,7 +290,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.OfferGroup.schema.strict.json b/gen/jsonschema/fora.v1.OfferGroup.schema.strict.json
index 8596ac08..3113fb06 100644
--- a/gen/jsonschema/fora.v1.OfferGroup.schema.strict.json
+++ b/gen/jsonschema/fora.v1.OfferGroup.schema.strict.json
@@ -264,7 +264,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -294,7 +294,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.ResourceResponse.schema.json b/gen/jsonschema/fora.v1.ResourceResponse.schema.json
index 8c810390..6eccf39e 100644
--- a/gen/jsonschema/fora.v1.ResourceResponse.schema.json
+++ b/gen/jsonschema/fora.v1.ResourceResponse.schema.json
@@ -260,7 +260,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -290,7 +290,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.ResourceResponse.schema.strict.json b/gen/jsonschema/fora.v1.ResourceResponse.schema.strict.json
index f118486b..ea485e45 100644
--- a/gen/jsonschema/fora.v1.ResourceResponse.schema.strict.json
+++ b/gen/jsonschema/fora.v1.ResourceResponse.schema.strict.json
@@ -264,7 +264,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -294,7 +294,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.TransactionItem.schema.json b/gen/jsonschema/fora.v1.TransactionItem.schema.json
index 46e62ae8..bede4a2d 100644
--- a/gen/jsonschema/fora.v1.TransactionItem.schema.json
+++ b/gen/jsonschema/fora.v1.TransactionItem.schema.json
@@ -280,7 +280,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -310,7 +310,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.TransactionItem.schema.strict.json b/gen/jsonschema/fora.v1.TransactionItem.schema.strict.json
index 1ec9e57b..0050310c 100644
--- a/gen/jsonschema/fora.v1.TransactionItem.schema.strict.json
+++ b/gen/jsonschema/fora.v1.TransactionItem.schema.strict.json
@@ -285,7 +285,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -315,7 +315,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.TransactionRequest.schema.json b/gen/jsonschema/fora.v1.TransactionRequest.schema.json
index ccbad2d7..04359d89 100644
--- a/gen/jsonschema/fora.v1.TransactionRequest.schema.json
+++ b/gen/jsonschema/fora.v1.TransactionRequest.schema.json
@@ -437,7 +437,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -467,7 +467,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/jsonschema/fora.v1.TransactionRequest.schema.strict.json b/gen/jsonschema/fora.v1.TransactionRequest.schema.strict.json
index 4a213a4a..55186eb7 100644
--- a/gen/jsonschema/fora.v1.TransactionRequest.schema.strict.json
+++ b/gen/jsonschema/fora.v1.TransactionRequest.schema.strict.json
@@ -446,7 +446,7 @@
"pricing": {
"$ref": "#/$defs/fora.v1.Pricing",
"description": "It derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.",
- "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
+ "title": "The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them."
},
"reporting": {
"$ref": "#/$defs/fora.v1.ReportingObligation",
@@ -476,7 +476,7 @@
"type": "array"
},
"terms": {
- "description": "The distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ "description": "The distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
"items": {
"$ref": "#/$defs/fora.v1.LicenseTerm"
},
diff --git a/gen/python/wire/models.py b/gen/python/wire/models.py
index 973286bc..153943a8 100644
--- a/gen/python/wire/models.py
+++ b/gen/python/wire/models.py
@@ -1948,7 +1948,7 @@ class Offer(WireModel):
)
pricing: Pricing | None = Field(
None,
- description='The offer\'s price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term\'s price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a "headline" picked among them.\n\nIt derives from exactly one catalog term: the term\'s pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer\'s price and its\n term\'s price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term\'s own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term\'s own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.',
+ description='The offer\'s price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term\'s price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a "headline" picked among them.\n\nIt derives from exactly one catalog term: the term\'s pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer\'s price and its\n term\'s price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term\'s own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term\'s own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.',
)
reporting: ReportingObligation | None = Field(
None,
@@ -1972,7 +1972,7 @@ class Offer(WireModel):
)
terms: list[LicenseTerm] | None = Field(
None,
- description="The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
+ description="The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.",
max_length=1,
min_length=1,
)
diff --git a/gen/ts/wire/schemas.ts b/gen/ts/wire/schemas.ts
index 253440f2..7d72c01a 100644
--- a/gen/ts/wire/schemas.ts
+++ b/gen/ts/wire/schemas.ts
@@ -50,7 +50,7 @@ export const DiscoveryMethodSchema = /* @__PURE__ */ (() => wire(z.enum(["DISCOV
export const DiscoveryRequestSchema = /* @__PURE__ */ (() => wire(z.object({ "acceptable_restrictions": z.array(z.object({ "axis": z.union([z.string().regex(new RegExp("^RESTRICTION_KIND_UNSPECIFIED$")), z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("Which axis (same enum as Restriction.kind): FUNCTION / GEOGRAPHY /\n USER_TYPE / OTHER.").default(0), "values": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("The values the query operates within on this axis — same token vocabulary\n as the terms (e.g. FUNCTION [\"ai-train\"], GEOGRAPHY [\"US\", \"EU\"]).").optional() }).describe("AcceptableRestriction — the limits a query operates within on one restriction\n axis, expressed in the same RestrictionKind vocabulary that terms use. The\n Exchange/Broker MAY pre-select offers whose term restrictions fall within\n these as a convenience (see Restriction); it is NOT enforcement — the agent\n self-selects and bears compliance.")).describe("The limits the agent will operate within, per restriction axis — see\n AcceptableRestriction. The Broker forwards these to Exchanges in\n ResourceQuery.acceptable_restrictions. Advisory selection inputs, not\n enforcement.").optional(), "constraints": z.object({ "budget_period": z.string().describe("Budget period (e.g. \"2592000s\" = 30 days; proto-JSON encodes Duration\n as seconds).").optional(), "budget_scope": z.string().describe("Budget scope identifier: whose budget this is.\n E.g. \"user:u-12345\" for per-user budgets, \"team:eng\" for per-team.").optional(), "delivery_preference": z.array(z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"])).describe("Preferred delivery methods, in order of preference.").optional(), "exchanges": z.array(z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260)).describe("Authorized Exchange domains, in the shape \"Request recipient\" defines in the\n file header. Broker queries only these. This is a FILTER over third parties,\n not an address — the recipient of the request carrying it is a separate\n question.").optional(), "max_data_age": z.string().describe("Maximum acceptable age of resource data. The Broker SHOULD\n exclude offers where (now - Offer.data_as_of) exceeds this duration.\n\nOnly relevant for DYNAMIC resources. Ignored for STATIC (content is\n immutable) and LIVE (content doesn't exist yet).\n\n Examples:\n 7 days — \"credit report updated within the last week\"\n 1 hour — \"stock snapshot from the last hour\"\n 30 days — \"drug interaction database updated this month\"").optional(), "max_hops": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Maximum number of hops the agent will allow its discovery to travel. It\n rides only on DiscoveryRequest, which terminates at the Broker, and a\n Broker originates its own queries rather than forwarding the agent's\n request, so no party receives or enforces it in this version: an Exchange\n never sees it, and an Exchange's own cap is\n WellKnownManifest.max_intermediary_hops. The field is kept for multi-hop\n Broker discovery, an open protocol question. Absent = no cap stated.").optional(), "max_price": z.object({ "amount": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Exact decimal string (not a float), e.g. \"19.99\". Denominated in `currency`.").default(""), "currency": z.string().default(""), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).optional() }).describe("Maximum price the agent is willing to pay.").optional(), "max_unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Maximum effective cost per unit, as an exact decimal string (not a float).").optional(), "period_budget": z.object({ "amount": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Exact decimal string (not a float), e.g. \"19.99\". Denominated in `currency`.").default(""), "currency": z.string().default(""), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).optional() }).describe("Per-period budget limit for the budget_scope.").optional(), "preferred_exchanges": z.array(z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260)).describe("Exchanges the agent has existing relationships with (subscriptions,\n contracts). The Broker SHOULD prefer these when resource is\n available — subscription resource has zero marginal cost.").optional(), "reporting_capable": z.boolean().describe("Whether the agent supports post-usage reporting.").optional() }).describe("Constraints for exchange filtering and offer selection.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "query": z.string().describe("Search query for Broker-side resource discovery.\n Used when the agent doesn't know specific URIs but wants the Broker\n to find matching resources across Exchanges.\n When present, the Broker interprets the query and discovers resources\n across Exchanges on the agent's behalf. Results returned as Offers\n in DiscoveryResponse, same as for specific URI requests.\n Can be used alongside uris (specific URIs + search in one request).").optional(), "requester": z.object({ "delegation": z.object({ "expires_at": z.string().datetime({ offset: true }).describe("Mirror of the chain's earliest `exp`. The Exchange MUST reject an expired\n token; the signed `exp` values decide, not this field.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "issuer": z.string().describe("Mirror of the ROOT token's `iss`, the identity configured with the trusted\n key (never an intermediate or terminal issuer). Informational: the\n Exchange never discovers a key from it, through OIDC, JWKS or otherwise.").optional(), "max_accesses": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Maximum number of accesses allowed under this delegation.\n Exchange tracks cumulative access count against this cap.\n Deny with DENIAL_REASON_QUOTA_EXCEEDED when count >= limit.\n For subscriptions with \"10,000 accesses/month\", this carries the ceiling.").optional(), "max_spend_cents": z.coerce.number().int().describe("Maximum spend in currency minor units (e.g., cents for USD).\n Exchange tracks cumulative spend against this cap.").optional(), "principal_domain": z.string().describe("Mirror of the ROOT token's `iss`, the configured trust anchor's identity.\n Informational: no key is looked up from it, and it establishes no trust.").default(""), "principal_id": z.string().describe("Principal's identifier (e.g., \"user@acme.com\", \"marketdata.example.com\").\n Informational; not read by the verifier.").default(""), "quota_period": z.string().describe("Quota reset period. How often the access/spend counters reset.\n Example: 30 days for monthly subscriptions — \"2592000s\" on the wire\n (proto-JSON encodes Duration as seconds; \"720h\" is not accepted).\n When absent, the quota is lifetime (bounded only by expires_at).").optional(), "revocation_uri": z.string().describe("Informational. Revocation lookup is not part of the delegation profile:\n the verifier never fetches this URI. A chain is invalidated early by\n withdrawing the configured root key.").optional(), "scopes": z.array(z.string()).describe("Mirror of the terminal token's effective scopes. In the chain, each\n child's scopes are covered by its parent's under the scope-matching rule\n on Requester.scopes (attenuation — can only narrow, not widen).").optional(), "token": z.string().regex(new RegExp("^[A-Za-z0-9+/]{0,43691}={0,2}$")).max(43692).describe("Token bytes. A JWT (base64url-encoded JWS). At most 32768 bytes: the\n complete chain, every embedded parent included, must fit. The limit is a\n wire rule, checked before the token is parsed and independent of whether\n the bytes are a valid JWT; a longer token is an invalid request, not a\n per-item denial.").default(""), "token_format": z.string().describe("Token format: \"jwt\" (default). Empty is treated as \"jwt\". The field stays\n open for a future format.").default("") }).describe("Optional delegation — present when the requester acts on behalf of\n another entity (user, organization, upstream agent). On a query a Broker\n originated it is not honoured, because the Broker signs alone and nothing\n on that leg can pass its holder binding; the Exchange answers as if none\n were present (see Delegation).").optional(), "domain": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the requester's key directory: the WBA directory at\n {domain}/.well-known/http-message-signatures-directory that publishes the\n agent's Ed25519 keys. It carries the bare-host shape \"Request recipient\"\n defines in the file header (a port allowed), for the same structural\n reason: a scheme, path or query smuggled in here would choose what gets\n fetched, not merely from where. It is never a free label. Every verifier\n reads it as the name of that directory, and the rule that binds it depends\n on whether the request is direct or relayed, which the receiver decides as\n \"Direct and relayed requests\" in the file header states:\n\n- Direct request. The agent's own RFC 9421 signature arrives because\n the agent sent the request itself. The verifier resolves the agent's\n keys from the COVERED `Signature-Agent` member, never from this field.\n It then MUST require this field to name that same directory: the host\n of the origin that member names, compared by\n the identity rule \"Request recipient\" defines (the shape check first,\n then case-folded, an absent port the same as \":443\", a subdomain a\n different party). A mismatch is refused as UNAUTHENTICATED with\n `request_auth_failure` SIGNATURE_INVALID: the signature verifies, but\n not for the requester the body names. A Broker applies this rule at\n BrokerService.ExecuteTransaction. At BrokerService.Resolve it refuses\n nothing for this field: every query it originates names the host of\n the origin that member names (see DiscoveryRequest.requester). An\n Exchange applies it on every request whose signer it does not admit as\n a Broker; a mismatch from a signer it does admit is a relayed request,\n one of the two cases below.\n - Purchase relayed through a Broker. BrokerService.ExecuteTransaction\n re-packages the purchase, so the request signature and the covered\n `Signature-Agent` member are the Broker's. They say only that the call comes\n from the Broker, and they sign no purchase. The Exchange MUST verify\n each item's AgentAcceptance, and the AgentRequestAcceptance, against\n the Ed25519 keys currently valid in the directory this field names\n (see AgentRequestAcceptance), never against the Broker's key. Those\n acceptances are the only agent signatures the Exchange sees, and the\n purchase belongs to the account of the directory they verify under\n (see \"Accounts\" in the file header).\n - Broker-led discovery. A ResourceQuery a Broker originated carries no agent\n signature, so nothing on that leg authenticates this field: it is the\n Broker's statement, under the Broker's own signature, of whom it\n queries for. A delegation on such a query is not honoured (see\n Delegation)."), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "id": z.string().min(1).max(255).describe("REQUIRED. A free label the agent chooses for attribution (e.g.\n \"agent-research-bot-001\"), 1 to 255 characters. Use it to tell apart the\n sub-agents or end customers that share one key directory. It is never\n identity: no verifier finds keys from it, and no party trusts it for an\n authentication or authorization decision. The requester's identity is\n `domain` together with the key that verifies. The acceptance signatures\n cover this label (see AgentAcceptancePayload), so a relaying Broker cannot\n rewrite it, but its value is still only what the agent says."), "name": z.string().describe("Human-readable name (e.g., \"Acme Research Assistant\").").optional(), "scopes": z.array(z.string()).max(64).describe("Entitlement scopes: the requester's own statement of what it is entitled\n to. On a query a Broker originated, they are the Broker's statement for the\n agent. They are signed and attributable, and the Exchange does not verify\n them. Scope coverage decides which terms the Exchange presents (see\n LicenseTerm.scopes, which also states when an Exchange MAY require a\n verified delegation); it is not access control. A false statement is\n traceable to the signing key, its directory and its account, and the\n agent's acceptance signs the scoped offer it buys; it is settled through\n reconciliation and the account relationship.\n\nEXISTENCE HIDING. A requester never learns about a resource outside its\n scopes. When the scopes leave no presentable offer for a resource, the\n Exchange answers with no offers and no absence reason (an OfferGroup with\n empty `offers` and `absence_reason` unset), exactly as for any resource\n with nothing to offer. A Broker relaying that answer adds no reason of its\n own. At purchase, an offer the Exchange presented is honoured until it\n expires, so a scope refusal cannot arise there. No reason in the protocol\n reports a scope shortfall: OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT and\n DENIAL_REASON_SCOPE_INSUFFICIENT are deprecated and never sent.\n\n Scope format: colon-separated segments, \"{domain}:{permission}\" or\n \"{profile}:{permission}\", optionally multi-segment (\"dist:US:CA\");\n matching is segment-wise per the rule below (no implicit hierarchy).\n Examples:\n \"credit:read\" — can access credit reports\n \"subscription:marketdata-2026\" — has active MarketData subscription\n \"academic:*\" — full access to academic resources\n \"internal:reports\" — can access internal reports\n \"*\" — unrestricted (public Exchange default)\n\n Matching is SEGMENT-WISE (\":\" separated). A granted scope G covers a\n required scope R iff, segment by segment, each G segment equals the\n corresponding R segment or is \"*\"; a terminal \"*\" matches all remaining\n segments, of which there must be at least one. There is NO implicit prefix\n match, and a grant NARROWER than the requirement does not cover it (G must\n be equal-to-or-broader than R), so a grant with more segments than the\n requirement never covers it, even when its last segment is \"*\". A \"*\" in R\n is a literal segment: a narrow grant never covers a wildcard requirement.\n Matching is case-sensitive and never trims. Examples: \"dist:*\" covers\n \"dist:US\", \"dist:US:CA\" and \"dist:US:*\" but not \"dist\"; \"dist:US:*\" covers\n \"dist:US:CA\" but not \"dist:EU\" or \"dist:US\"; \"publisher:*:read\" covers\n \"publisher:news:read\" but not \"publisher:news:*\"; bare \"dist\" covers only\n \"dist\"; granted \"dist:US:CA\" does NOT cover required \"dist:US\"; \"*\" covers\n everything. This same rule governs LicenseTerm.scopes and the attenuation\n of a delegation child against its parent — one algorithm protocol-wide.\n\n When empty, Exchange applies its default access policy (typically\n returns all publicly available resources).").optional(), "type": z.enum(["REQUESTER_TYPE_AGENT","REQUESTER_TYPE_HUMAN_TOOL","REQUESTER_TYPE_SERVICE","REQUESTER_TYPE_DELEGATED","REQUESTER_TYPE_RESEARCH"]).describe("What kind of entity is making this request.") }).describe("Requester identity — who is making this request, the scopes it states.\n The Broker forwards this to Exchanges in ResourceQuery.requester, with\n Requester.domain the host of the origin that the covered `Signature-Agent`\n member of the agent's verified signature names. A Requester.domain that\n names another host is never forwarded, and the Broker answers the request\n all the same (see \"Direct and relayed requests\" in the file header).\n REQUIRED: a discovery names its requester, whose Requester.id and\n Requester.domain are required in turn."), "search_filters": z.record(z.string(), z.any()).describe("Structured search filters (optional, alongside or instead of query).\n Keys are profile-specific: \"academic.topic\", \"news.category\",\n \"legal.jurisdiction\", etc. The Broker maps these to Exchange-specific\n query parameters.").optional(), "supported_profiles": z.array(z.string()).describe("Domain extension profiles the agent understands.\n\nThe Broker MUST forward this list unchanged in\n ResourceQuery.supported_profiles on every query it sends an Exchange, and\n forward an absent list as absent. It does not route by it: which Exchanges\n it queries is decided on other grounds, and each Exchange answers for its\n own profiles (see ResourceQuery.supported_profiles).\n\n Examples: [\"fora-academic-v1\"] — agent working on literature review").optional(), "uris": z.array(z.string()).max(256).describe("Resource URIs the agent wants. The Broker forwards these to Exchanges in\n ResourceQuery.uris. Optional when `query` / `search_filters` drive\n Broker-side discovery instead.").optional(), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("DiscoveryRequest — Agent sends to Broker (Step 1).")))();
-export const DiscoveryResponseSchema = /* @__PURE__ */ (() => wire(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Existence-oracle note: an authorization-flavored reason (NOT_AUTHORIZED,\n NOT_IN_CATALOG, CONTENT_BLOCKED) confirms a resource exists\n and why access was refused. Resolve surfaces the same oracle at the broker\n that OfferGroup.absence_reason does at the Exchange, so the same mitigation\n applies: where existence itself must stay hidden, the Broker MAY omit the\n reason (leave this unset) rather than reveal it.\n A scope shortfall never reaches this field: the Exchange answers it with\n no absence reason, and the Broker MUST NOT supply one (the existence-hiding\n rule, stated on Requester.scopes).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "offer_groups": z.array(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Why no offers are available for this URI.\n Set only when `offers` is empty, and not always then: a resource the\n requester's scopes leave with no presentable offer is answered with empty\n `offers` and this field unset (the existence-hiding rule, stated on\n Requester.scopes). Enables agents/Brokers to distinguish\n \"resource not in catalog\" from \"resource blocked for your use case\" without\n trial-and-error transactions. Analogous to OpenRTB nbr codes and\n Shutterstock per-item error metadata in batch responses.").optional(), "discovery_method": z.enum(["DISCOVERY_METHOD_EXCHANGE","DISCOVERY_METHOD_SEARCH","DISCOVERY_METHOD_RECOMMENDATION","DISCOVERY_METHOD_SYNDICATION"]).describe("How this URI was discovered by the Broker (v2 extension point).\n v1: always DISCOVERY_METHOD_EXCHANGE (Broker queried an Exchange).\n v2: may include DISCOVERY_METHOD_SEARCH (URI found via search engine like Exa),\n DISCOVERY_METHOD_RECOMMENDATION, etc. The Broker discovers URIs\n through any source, then routes through Exchange for pricing/transaction.\n The discovery method does not affect the transaction flow — it's metadata\n for the agent to understand how the resource was found.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Zero or more offers for this URI. Empty = resource not available.").optional(), "restriction_filters": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("When absence_reason = RESTRICTION_FILTERED, the restriction axes that drove\n the convenience pre-filter, in the same RestrictionKind vocabulary the terms\n use (e.g. [GEOGRAPHY] when the requester's stated geography matched no term).\n Advisory diagnostics, not an enforcement verdict.").optional(), "uri": z.string().describe("The URI this group of offers is for (echoed from ResourceQuery.uris).").default("") }).describe("OfferGroup — Offers for a single requested URI.\n Enables multi-URI batch queries where the caller needs to know\n which offers correspond to which requested resource.")).describe("Offers grouped by requested URI — the sole offer representation in this\n response. One OfferGroup per URI the agent asked for (echoed in\n OfferGroup.uri); a group with no offers carries OfferGroup.absence_reason\n explaining why. Each contained Offer is the full signed Offer the Exchange\n issued (including Offer.exchange, the execute-routing target), forwarded by\n the Broker unchanged so the agent can verify the signature end to end.").optional(), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("DiscoveryResponse — Broker returns to Agent (Step 6).\n\nCarries discovery results only: the offers the Broker gathered across\n Exchanges, grouped by the URI they were requested for. Committing to an offer\n is a separate exchange on the execute path; that per-transaction result\n (transaction_id, billing_id, cost, delivery_method, retrieval endpoint, …)\n is returned by TransactionResponse, or by BrokerTransactionResponse when the\n purchase goes through BrokerService.ExecuteTransaction, not here.")))();
+export const DiscoveryResponseSchema = /* @__PURE__ */ (() => wire(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Existence-oracle note: an authorization-flavored reason (NOT_AUTHORIZED,\n NOT_IN_CATALOG, CONTENT_BLOCKED) confirms a resource exists\n and why access was refused. Resolve surfaces the same oracle at the broker\n that OfferGroup.absence_reason does at the Exchange, so the same mitigation\n applies: where existence itself must stay hidden, the Broker MAY omit the\n reason (leave this unset) rather than reveal it.\n A scope shortfall never reaches this field: the Exchange answers it with\n no absence reason, and the Broker MUST NOT supply one (the existence-hiding\n rule, stated on Requester.scopes).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "offer_groups": z.array(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Why no offers are available for this URI.\n Set only when `offers` is empty, and not always then: a resource the\n requester's scopes leave with no presentable offer is answered with empty\n `offers` and this field unset (the existence-hiding rule, stated on\n Requester.scopes). Enables agents/Brokers to distinguish\n \"resource not in catalog\" from \"resource blocked for your use case\" without\n trial-and-error transactions. Analogous to OpenRTB nbr codes and\n Shutterstock per-item error metadata in batch responses.").optional(), "discovery_method": z.enum(["DISCOVERY_METHOD_EXCHANGE","DISCOVERY_METHOD_SEARCH","DISCOVERY_METHOD_RECOMMENDATION","DISCOVERY_METHOD_SYNDICATION"]).describe("How this URI was discovered by the Broker (v2 extension point).\n v1: always DISCOVERY_METHOD_EXCHANGE (Broker queried an Exchange).\n v2: may include DISCOVERY_METHOD_SEARCH (URI found via search engine like Exa),\n DISCOVERY_METHOD_RECOMMENDATION, etc. The Broker discovers URIs\n through any source, then routes through Exchange for pricing/transaction.\n The discovery method does not affect the transaction flow — it's metadata\n for the agent to understand how the resource was found.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Zero or more offers for this URI. Empty = resource not available.").optional(), "restriction_filters": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("When absence_reason = RESTRICTION_FILTERED, the restriction axes that drove\n the convenience pre-filter, in the same RestrictionKind vocabulary the terms\n use (e.g. [GEOGRAPHY] when the requester's stated geography matched no term).\n Advisory diagnostics, not an enforcement verdict.").optional(), "uri": z.string().describe("The URI this group of offers is for (echoed from ResourceQuery.uris).").default("") }).describe("OfferGroup — Offers for a single requested URI.\n Enables multi-URI batch queries where the caller needs to know\n which offers correspond to which requested resource.")).describe("Offers grouped by requested URI — the sole offer representation in this\n response. One OfferGroup per URI the agent asked for (echoed in\n OfferGroup.uri); a group with no offers carries OfferGroup.absence_reason\n explaining why. Each contained Offer is the full signed Offer the Exchange\n issued (including Offer.exchange, the execute-routing target), forwarded by\n the Broker unchanged so the agent can verify the signature end to end.").optional(), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("DiscoveryResponse — Broker returns to Agent (Step 6).\n\nCarries discovery results only: the offers the Broker gathered across\n Exchanges, grouped by the URI they were requested for. Committing to an offer\n is a separate exchange on the execute path; that per-transaction result\n (transaction_id, billing_id, cost, delivery_method, retrieval endpoint, …)\n is returned by TransactionResponse, or by BrokerTransactionResponse when the\n purchase goes through BrokerService.ExecuteTransaction, not here.")))();
export const DisputeFailureSchema = /* @__PURE__ */ (() => wire(z.object({ "reason": z.enum(["DISPUTE_FAILURE_REASON_TRANSACTION_NOT_FOUND","DISPUTE_FAILURE_REASON_REPORT_NOT_FILED","DISPUTE_FAILURE_REASON_WINDOW_EXPIRED","DISPUTE_FAILURE_REASON_DUPLICATE","DISPUTE_FAILURE_REASON_INELIGIBLE","DISPUTE_FAILURE_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("The failure reason (defined-only, non-zero)") }).describe("DisputeFailure — a dispute could not be filed.")))();
@@ -100,11 +100,11 @@ export const ObligationKindSchema = /* @__PURE__ */ (() => wire(z.enum(["OBLIGAT
export const ObligationTriggerSchema = /* @__PURE__ */ (() => wire(z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"])))();
-export const OfferSchema = /* @__PURE__ */ (() => wire(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")))();
+export const OfferSchema = /* @__PURE__ */ (() => wire(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")))();
export const OfferAbsenceReasonSchema = /* @__PURE__ */ (() => wire(z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"])))();
-export const OfferGroupSchema = /* @__PURE__ */ (() => wire(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Why no offers are available for this URI.\n Set only when `offers` is empty, and not always then: a resource the\n requester's scopes leave with no presentable offer is answered with empty\n `offers` and this field unset (the existence-hiding rule, stated on\n Requester.scopes). Enables agents/Brokers to distinguish\n \"resource not in catalog\" from \"resource blocked for your use case\" without\n trial-and-error transactions. Analogous to OpenRTB nbr codes and\n Shutterstock per-item error metadata in batch responses.").optional(), "discovery_method": z.enum(["DISCOVERY_METHOD_EXCHANGE","DISCOVERY_METHOD_SEARCH","DISCOVERY_METHOD_RECOMMENDATION","DISCOVERY_METHOD_SYNDICATION"]).describe("How this URI was discovered by the Broker (v2 extension point).\n v1: always DISCOVERY_METHOD_EXCHANGE (Broker queried an Exchange).\n v2: may include DISCOVERY_METHOD_SEARCH (URI found via search engine like Exa),\n DISCOVERY_METHOD_RECOMMENDATION, etc. The Broker discovers URIs\n through any source, then routes through Exchange for pricing/transaction.\n The discovery method does not affect the transaction flow — it's metadata\n for the agent to understand how the resource was found.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Zero or more offers for this URI. Empty = resource not available.").optional(), "restriction_filters": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("When absence_reason = RESTRICTION_FILTERED, the restriction axes that drove\n the convenience pre-filter, in the same RestrictionKind vocabulary the terms\n use (e.g. [GEOGRAPHY] when the requester's stated geography matched no term).\n Advisory diagnostics, not an enforcement verdict.").optional(), "uri": z.string().describe("The URI this group of offers is for (echoed from ResourceQuery.uris).").default("") }).describe("OfferGroup — Offers for a single requested URI.\n Enables multi-URI batch queries where the caller needs to know\n which offers correspond to which requested resource.")))();
+export const OfferGroupSchema = /* @__PURE__ */ (() => wire(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Why no offers are available for this URI.\n Set only when `offers` is empty, and not always then: a resource the\n requester's scopes leave with no presentable offer is answered with empty\n `offers` and this field unset (the existence-hiding rule, stated on\n Requester.scopes). Enables agents/Brokers to distinguish\n \"resource not in catalog\" from \"resource blocked for your use case\" without\n trial-and-error transactions. Analogous to OpenRTB nbr codes and\n Shutterstock per-item error metadata in batch responses.").optional(), "discovery_method": z.enum(["DISCOVERY_METHOD_EXCHANGE","DISCOVERY_METHOD_SEARCH","DISCOVERY_METHOD_RECOMMENDATION","DISCOVERY_METHOD_SYNDICATION"]).describe("How this URI was discovered by the Broker (v2 extension point).\n v1: always DISCOVERY_METHOD_EXCHANGE (Broker queried an Exchange).\n v2: may include DISCOVERY_METHOD_SEARCH (URI found via search engine like Exa),\n DISCOVERY_METHOD_RECOMMENDATION, etc. The Broker discovers URIs\n through any source, then routes through Exchange for pricing/transaction.\n The discovery method does not affect the transaction flow — it's metadata\n for the agent to understand how the resource was found.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Zero or more offers for this URI. Empty = resource not available.").optional(), "restriction_filters": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("When absence_reason = RESTRICTION_FILTERED, the restriction axes that drove\n the convenience pre-filter, in the same RestrictionKind vocabulary the terms\n use (e.g. [GEOGRAPHY] when the requester's stated geography matched no term).\n Advisory diagnostics, not an enforcement verdict.").optional(), "uri": z.string().describe("The URI this group of offers is for (echoed from ResourceQuery.uris).").default("") }).describe("OfferGroup — Offers for a single requested URI.\n Enables multi-URI batch queries where the caller needs to know\n which offers correspond to which requested resource.")))();
export const PreviewSchema = /* @__PURE__ */ (() => wire(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")))();
@@ -172,7 +172,7 @@ export const ResourceQuerySchema = /* @__PURE__ */ (() => wire(z.object({ "accep
export const ResourceRefSchema = /* @__PURE__ */ (() => wire(z.object({ "domain": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("The bare host the resource lives on, as ResourceEntry.domain."), "path": z.string().regex(new RegExp("^/[^?#\\x00-\\x20\\x7f]*$")).min(1).max(2048).describe("The resource's absolute path, as ResourceEntry.path.") }).describe("ResourceRef — one catalog resource, named the way PushResources names it:\n by ResourceEntry.domain and ResourceEntry.path, under the same rules, which\n together form the catalog URI.")))();
-export const ResourceResponseSchema = /* @__PURE__ */ (() => wire(z.object({ "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("Canonical domain of the responding Exchange, in the shape \"Request\n recipient\" defines in the file header. The response counterpart of the\n recipient field on the request: it names who answered."), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "offer_groups": z.array(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Why no offers are available for this URI.\n Set only when `offers` is empty, and not always then: a resource the\n requester's scopes leave with no presentable offer is answered with empty\n `offers` and this field unset (the existence-hiding rule, stated on\n Requester.scopes). Enables agents/Brokers to distinguish\n \"resource not in catalog\" from \"resource blocked for your use case\" without\n trial-and-error transactions. Analogous to OpenRTB nbr codes and\n Shutterstock per-item error metadata in batch responses.").optional(), "discovery_method": z.enum(["DISCOVERY_METHOD_EXCHANGE","DISCOVERY_METHOD_SEARCH","DISCOVERY_METHOD_RECOMMENDATION","DISCOVERY_METHOD_SYNDICATION"]).describe("How this URI was discovered by the Broker (v2 extension point).\n v1: always DISCOVERY_METHOD_EXCHANGE (Broker queried an Exchange).\n v2: may include DISCOVERY_METHOD_SEARCH (URI found via search engine like Exa),\n DISCOVERY_METHOD_RECOMMENDATION, etc. The Broker discovers URIs\n through any source, then routes through Exchange for pricing/transaction.\n The discovery method does not affect the transaction flow — it's metadata\n for the agent to understand how the resource was found.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Zero or more offers for this URI. Empty = resource not available.").optional(), "restriction_filters": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("When absence_reason = RESTRICTION_FILTERED, the restriction axes that drove\n the convenience pre-filter, in the same RestrictionKind vocabulary the terms\n use (e.g. [GEOGRAPHY] when the requester's stated geography matched no term).\n Advisory diagnostics, not an enforcement verdict.").optional(), "uri": z.string().describe("The URI this group of offers is for (echoed from ResourceQuery.uris).").default("") }).describe("OfferGroup — Offers for a single requested URI.\n Enables multi-URI batch queries where the caller needs to know\n which offers correspond to which requested resource.")).describe("Offers grouped by requested URI (for multi-URI batch queries).\n When populated, `offers` SHOULD be empty to avoid ambiguity.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Flat list of offers (for single-URI queries).").optional(), "rate_limit": z.object({ "limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Maximum requests allowed in the current window.").optional(), "remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Requests remaining in the current window.").optional(), "reset_at": z.string().datetime({ offset: true }).describe("When the current window resets (UTC). After this time, `remaining` resets to `limit`.").optional(), "window": z.string().describe("Duration of the rate limit window (e.g. 60s = per-minute limit).").optional() }).describe("Rate limit status for this caller.\n Present when the Exchange enforces per-caller rate limits on discovery.\n Enables agents/Brokers to throttle proactively rather than hitting\n hard limits. Particularly important when a Broker fans out the\n same batch query to multiple Exchanges — mid-batch rate limiting\n can cause partial results if not signaled early.").optional(), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("ResourceResponse — Exchange returns candidate resource offers.\n\nWhen the ResourceQuery contains multiple URIs, offers are grouped by URI\n via OfferGroup. When a single URI is queried, the Exchange MAY use\n either the flat `offers` field or a single OfferGroup.")))();
+export const ResourceResponseSchema = /* @__PURE__ */ (() => wire(z.object({ "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("Canonical domain of the responding Exchange, in the shape \"Request\n recipient\" defines in the file header. The response counterpart of the\n recipient field on the request: it names who answered."), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "offer_groups": z.array(z.object({ "absence_reason": z.enum(["OFFER_ABSENCE_REASON_NOT_IN_CATALOG","OFFER_ABSENCE_REASON_CONTENT_BLOCKED","OFFER_ABSENCE_REASON_RESTRICTION_FILTERED","OFFER_ABSENCE_REASON_TEMPORARILY_UNAVAILABLE","OFFER_ABSENCE_REASON_NOT_AUTHORIZED","OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT","OFFER_ABSENCE_REASON_UNKNOWN_CRITICAL_EXTENSION","OFFER_ABSENCE_REASON_BUDGET_EXCEEDED"]).describe("Why no offers are available for this URI.\n Set only when `offers` is empty, and not always then: a resource the\n requester's scopes leave with no presentable offer is answered with empty\n `offers` and this field unset (the existence-hiding rule, stated on\n Requester.scopes). Enables agents/Brokers to distinguish\n \"resource not in catalog\" from \"resource blocked for your use case\" without\n trial-and-error transactions. Analogous to OpenRTB nbr codes and\n Shutterstock per-item error metadata in batch responses.").optional(), "discovery_method": z.enum(["DISCOVERY_METHOD_EXCHANGE","DISCOVERY_METHOD_SEARCH","DISCOVERY_METHOD_RECOMMENDATION","DISCOVERY_METHOD_SYNDICATION"]).describe("How this URI was discovered by the Broker (v2 extension point).\n v1: always DISCOVERY_METHOD_EXCHANGE (Broker queried an Exchange).\n v2: may include DISCOVERY_METHOD_SEARCH (URI found via search engine like Exa),\n DISCOVERY_METHOD_RECOMMENDATION, etc. The Broker discovers URIs\n through any source, then routes through Exchange for pricing/transaction.\n The discovery method does not affect the transaction flow — it's metadata\n for the agent to understand how the resource was found.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Zero or more offers for this URI. Empty = resource not available.").optional(), "restriction_filters": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("When absence_reason = RESTRICTION_FILTERED, the restriction axes that drove\n the convenience pre-filter, in the same RestrictionKind vocabulary the terms\n use (e.g. [GEOGRAPHY] when the requester's stated geography matched no term).\n Advisory diagnostics, not an enforcement verdict.").optional(), "uri": z.string().describe("The URI this group of offers is for (echoed from ResourceQuery.uris).").default("") }).describe("OfferGroup — Offers for a single requested URI.\n Enables multi-URI batch queries where the caller needs to know\n which offers correspond to which requested resource.")).describe("Offers grouped by requested URI (for multi-URI batch queries).\n When populated, `offers` SHOULD be empty to avoid ambiguity.").optional(), "offers": z.array(z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("Offer — A single resource offer from an Exchange.\n\nCombines pricing, delivery method, resource identity, and reporting terms.\n CoMP-specific metadata (Package, Function) available via fora-comp-v1 extension profile.")).describe("Flat list of offers (for single-URI queries).").optional(), "rate_limit": z.object({ "limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Maximum requests allowed in the current window.").optional(), "remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Requests remaining in the current window.").optional(), "reset_at": z.string().datetime({ offset: true }).describe("When the current window resets (UTC). After this time, `remaining` resets to `limit`.").optional(), "window": z.string().describe("Duration of the rate limit window (e.g. 60s = per-minute limit).").optional() }).describe("Rate limit status for this caller.\n Present when the Exchange enforces per-caller rate limits on discovery.\n Enables agents/Brokers to throttle proactively rather than hitting\n hard limits. Particularly important when a Broker fans out the\n same batch query to multiple Exchanges — mid-batch rate limiting\n can cause partial results if not signaled early.").optional(), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("ResourceResponse — Exchange returns candidate resource offers.\n\nWhen the ResourceQuery contains multiple URIs, offers are grouped by URI\n via OfferGroup. When a single URI is queried, the Exchange MAY use\n either the flat `offers` field or a single OfferGroup.")))();
export const RestrictionSchema = /* @__PURE__ */ (() => wire(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")))();
@@ -200,9 +200,9 @@ export const TermSemanticsSchema = /* @__PURE__ */ (() => wire(z.enum(["TERM_SEM
export const TransactionDenialSchema = /* @__PURE__ */ (() => wire(z.object({ "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("Bare host of the Exchange that PRODUCED this denial, in the form \"Request\n recipient\" defines in the file header. Not an echo of what the caller sent:\n on a relayed or fanned-out execute the request went to a Broker, so the\n Exchange that refused may not be one the agent named (through\n BrokerService.ExecuteTransaction this detail reaches the agent inside\n UpstreamRefusal.detail, beside UpstreamRefusal.party). Carrying it here is\n what lets ACCOUNT_NOT_REGISTERED be actionable — the agent learns where to\n call Register without fetching a manifest to work it out. NOTHING SIGNS THIS\n VALUE: it rides in a response, and on a relayed path the response passed\n through an intermediary, so this field is exactly the unsigned addressing\n the request-side `exchange` field exists to refuse. Treat it as a HINT, not\n an instruction. Before acting on it — and registering is a consequential act,\n handing an operator's business data and a signed acceptance of that\n Exchange's terms to whoever answers — a caller MUST check the value against\n a domain it already trusts for this transaction: the signed `offer.exchange`\n of the items in the refused request, or its own RequestConstraints.exchanges\n set. A value matching neither is reported to the caller and never dialled,\n because a hostile intermediary that could choose it would be choosing where\n an unattended agent registers.").optional(), "offer_id": z.string().describe("DEPRECATED, never set. A whole-request refusal pertains to no one offer; a\n denial of one offer rides on that offer's TransactionResultItem. The field\n is retained because removing it would break the v1 wire contract; a\n receiver ignores it.").optional(), "reason": z.enum(["DENIAL_REASON_ACCOUNT_INACTIVE","DENIAL_REASON_INSUFFICIENT_BALANCE","DENIAL_REASON_RATE_LIMITED","DENIAL_REASON_CONTENT_UNAVAILABLE","DENIAL_REASON_RESTRICTION_NOT_SATISFIED","DENIAL_REASON_REPORTING_OVERDUE","DENIAL_REASON_OFFER_EXPIRED","DENIAL_REASON_SIGNATURE_INVALID","DENIAL_REASON_QUOTA_EXCEEDED","DENIAL_REASON_DELEGATION_INVALID","DENIAL_REASON_SCOPE_INSUFFICIENT","DENIAL_REASON_ENTITLEMENT_MISSING","DENIAL_REASON_ENTITLEMENT_MALFORMED","DENIAL_REASON_ENTITLEMENT_EXPIRED","DENIAL_REASON_ENTITLEMENT_WRONG_BUYER","DENIAL_REASON_SUBSCRIPTION_LAPSED","DENIAL_REASON_ENTITLEMENT_NOT_GRANTED","DENIAL_REASON_ACCOUNT_NOT_REGISTERED","DENIAL_REASON_RELAY_NOT_ACCEPTED","DENIAL_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("The denial reason (defined-only, non-zero)"), "restriction_mismatches": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("DEPRECATED, never set. RESTRICTION_NOT_SATISFIED is never sent: the\n Exchange never enforces a restriction (see Restriction). The field is\n retained because removing it would break the v1 wire contract; a receiver\n ignores it.").optional() }).describe("TransactionDenial — ExecuteTransaction refused the WHOLE request. Reuses the\n DenialReason vocabulary.\n\nThis is the only case an ExecuteTransaction answer is a non-OK error carrying\n a DenialReason: the Exchange refuses the request as a whole and decides no\n item, for a reason about the requester or the request that would deny every\n item alike — for example the requester holds no account\n (ACCOUNT_NOT_REGISTERED), its account is not active (ACCOUNT_INACTIVE), it has\n overdue reports (REPORTING_OVERDUE), or the call is rate limited\n (RATE_LIMITED). On a purchase a Broker relayed, the requester is the agent and\n the account is the agent's, never the Broker's (see \"Accounts\" in the file\n header). Nothing is\n purchased. Every decision about an item is answered in the body instead, on\n TransactionResultItem.denial_reason, whatever the item count: a denied\n one-item purchase is a successful response whose one item is denied, never\n this error. RELAY_NOT_ACCEPTED is a per-item decision and is never carried\n here, and RESTRICTION_NOT_SATISFIED is never sent at all.")))();
-export const TransactionItemSchema = /* @__PURE__ */ (() => wire(z.object({ "agent_acceptance": z.object({ "signature": z.string().min(1).describe("Hex-encoded detached Ed25519 signature over the canonical AgentAcceptancePayload\n bytes (see the canonical-signing definition on Offer.signature)."), "signature_algorithm": z.string().describe("Signature algorithm; \"EdDSA\" for Ed25519.").default("") }).describe("The agent's detached acceptance signature over this item's `offer`.\n Optional on the wire; the Exchange enforces presence per\n item at the service layer for relayed batches. Signed bytes = the canonical\n AgentAcceptancePayload form, with requester_* and idempotency_key\n taken from the ENCLOSING TransactionRequest and offer_sig = offer.signature.").optional(), "offer": z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("The FULL signed Offer for this batch entry, reflected back exactly as\n received at discovery. The Exchange verifies `offer.signature` over these\n presented bytes — stateless, no reconstruct-from-catalog. REQUIRED: every\n batch item carries its offer.") }).describe("TransactionItem — A single offer commitment within a batch transaction.")))();
+export const TransactionItemSchema = /* @__PURE__ */ (() => wire(z.object({ "agent_acceptance": z.object({ "signature": z.string().min(1).describe("Hex-encoded detached Ed25519 signature over the canonical AgentAcceptancePayload\n bytes (see the canonical-signing definition on Offer.signature)."), "signature_algorithm": z.string().describe("Signature algorithm; \"EdDSA\" for Ed25519.").default("") }).describe("The agent's detached acceptance signature over this item's `offer`.\n Optional on the wire; the Exchange enforces presence per\n item at the service layer for relayed batches. Signed bytes = the canonical\n AgentAcceptancePayload form, with requester_* and idempotency_key\n taken from the ENCLOSING TransactionRequest and offer_sig = offer.signature.").optional(), "offer": z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("The FULL signed Offer for this batch entry, reflected back exactly as\n received at discovery. The Exchange verifies `offer.signature` over these\n presented bytes — stateless, no reconstruct-from-catalog. REQUIRED: every\n batch item carries its offer.") }).describe("TransactionItem — A single offer commitment within a batch transaction.")))();
-export const TransactionRequestSchema = /* @__PURE__ */ (() => wire(z.object({ "agent_request_acceptance": z.object({ "payload": z.object({ "idempotency_key": z.string().default(""), "items": z.array(z.object({ "exchange": z.string().min(1), "offer_sig": z.string().min(1) }).describe("AgentRequestAcceptanceItem is the minimum reference needed to authorize an\n offer's membership, order, and fan-out destination without repeating the\n full Offer in every projected subrequest. Offer.signature transitively binds\n the full offer, including its exchange field; the explicit exchange lets a\n recipient derive which signed references must appear in its projection.")).min(1).max(256).describe("Complete original request order, before Broker fan-out. Capped at 256 —\n the same ceiling a discovery query's uris list carries, so one request\n can reference at most one offer per queried URI at the query cap. The Go\n verification helper enforces the same bound itself before doing any\n canonicalization work, because a verifier may run with wire validation\n off and the canonical rendering of an unbounded list is the expensive\n step an unauthenticated caller could otherwise buy for free.").optional(), "requester_domain": z.string().describe("Requester.domain of the request. Non-empty (see AgentRequestAcceptance).").default(""), "requester_id": z.string().describe("Requester.id of the request. Non-empty (see AgentRequestAcceptance).").default("") }).describe("The signed payload is carried because a projected subrequest does not carry\n offers addressed to other Exchanges and therefore cannot reconstruct the\n original complete set by itself."), "signature": z.string().min(1).describe("Hex-encoded detached Ed25519 signature over the canonical payload bytes."), "signature_algorithm": z.string().describe("Signature algorithm; \"EdDSA\" for Ed25519.").default("") }).describe("Optional for wire compatibility. When present, an Exchange verifies this\n before creating or serving request-level idempotency state. A Broker MUST\n forward it unchanged on every projected subrequest. Older clients that omit\n it retain per-item execution semantics but receive no request-level claim.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "idempotency_key": z.string().min(1).max(255).describe("Idempotency key (REQUIRED). The server MUST dedupe on this: a replay returns\n the original result rather than re-executing, and the same key with\n different items is refused with already_exists. The key is scoped to the\n agent, whatever the route: a purchase repeated directly or through any\n Broker the Exchange admits is the same purchase, and the Broker forwards the\n agent's key unchanged. A key another agent or tenant used is never a\n collision. The full rule is \"Idempotency\" in the file header. The\n transaction's durable identity is the Exchange-assigned transaction_id in\n the response."), "items": z.array(z.object({ "agent_acceptance": z.object({ "signature": z.string().min(1).describe("Hex-encoded detached Ed25519 signature over the canonical AgentAcceptancePayload\n bytes (see the canonical-signing definition on Offer.signature)."), "signature_algorithm": z.string().describe("Signature algorithm; \"EdDSA\" for Ed25519.").default("") }).describe("The agent's detached acceptance signature over this item's `offer`.\n Optional on the wire; the Exchange enforces presence per\n item at the service layer for relayed batches. Signed bytes = the canonical\n AgentAcceptancePayload form, with requester_* and idempotency_key\n taken from the ENCLOSING TransactionRequest and offer_sig = offer.signature.").optional(), "offer": z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price a Broker ranks and compares across\n Exchanges, and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what a Broker ranks on and\n what execute charges, so a term that is not on its own offer has no price\n a Broker can compare and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("The FULL signed Offer for this batch entry, reflected back exactly as\n received at discovery. The Exchange verifies `offer.signature` over these\n presented bytes — stateless, no reconstruct-from-catalog. REQUIRED: every\n batch item carries its offer.") }).describe("TransactionItem — A single offer commitment within a batch transaction.")).min(1).describe("The offers committed in this request (REQUIRED, min 1), each carrying its\n own reflected signed Offer + detached acceptance. A single offer is the\n degenerate 1-element list. The Exchange verifies each item's\n `offer.signature` (which covers pricing, terms, and expires_at) over the\n presented bytes against its own key — stateless, self-contained bearer\n tokens, with no reconstruct-from-catalog.").optional(), "requester": z.object({ "delegation": z.object({ "expires_at": z.string().datetime({ offset: true }).describe("Mirror of the chain's earliest `exp`. The Exchange MUST reject an expired\n token; the signed `exp` values decide, not this field.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "issuer": z.string().describe("Mirror of the ROOT token's `iss`, the identity configured with the trusted\n key (never an intermediate or terminal issuer). Informational: the\n Exchange never discovers a key from it, through OIDC, JWKS or otherwise.").optional(), "max_accesses": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Maximum number of accesses allowed under this delegation.\n Exchange tracks cumulative access count against this cap.\n Deny with DENIAL_REASON_QUOTA_EXCEEDED when count >= limit.\n For subscriptions with \"10,000 accesses/month\", this carries the ceiling.").optional(), "max_spend_cents": z.coerce.number().int().describe("Maximum spend in currency minor units (e.g., cents for USD).\n Exchange tracks cumulative spend against this cap.").optional(), "principal_domain": z.string().describe("Mirror of the ROOT token's `iss`, the configured trust anchor's identity.\n Informational: no key is looked up from it, and it establishes no trust.").default(""), "principal_id": z.string().describe("Principal's identifier (e.g., \"user@acme.com\", \"marketdata.example.com\").\n Informational; not read by the verifier.").default(""), "quota_period": z.string().describe("Quota reset period. How often the access/spend counters reset.\n Example: 30 days for monthly subscriptions — \"2592000s\" on the wire\n (proto-JSON encodes Duration as seconds; \"720h\" is not accepted).\n When absent, the quota is lifetime (bounded only by expires_at).").optional(), "revocation_uri": z.string().describe("Informational. Revocation lookup is not part of the delegation profile:\n the verifier never fetches this URI. A chain is invalidated early by\n withdrawing the configured root key.").optional(), "scopes": z.array(z.string()).describe("Mirror of the terminal token's effective scopes. In the chain, each\n child's scopes are covered by its parent's under the scope-matching rule\n on Requester.scopes (attenuation — can only narrow, not widen).").optional(), "token": z.string().regex(new RegExp("^[A-Za-z0-9+/]{0,43691}={0,2}$")).max(43692).describe("Token bytes. A JWT (base64url-encoded JWS). At most 32768 bytes: the\n complete chain, every embedded parent included, must fit. The limit is a\n wire rule, checked before the token is parsed and independent of whether\n the bytes are a valid JWT; a longer token is an invalid request, not a\n per-item denial.").default(""), "token_format": z.string().describe("Token format: \"jwt\" (default). Empty is treated as \"jwt\". The field stays\n open for a future format.").default("") }).describe("Optional delegation — present when the requester acts on behalf of\n another entity (user, organization, upstream agent). On a query a Broker\n originated it is not honoured, because the Broker signs alone and nothing\n on that leg can pass its holder binding; the Exchange answers as if none\n were present (see Delegation).").optional(), "domain": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the requester's key directory: the WBA directory at\n {domain}/.well-known/http-message-signatures-directory that publishes the\n agent's Ed25519 keys. It carries the bare-host shape \"Request recipient\"\n defines in the file header (a port allowed), for the same structural\n reason: a scheme, path or query smuggled in here would choose what gets\n fetched, not merely from where. It is never a free label. Every verifier\n reads it as the name of that directory, and the rule that binds it depends\n on whether the request is direct or relayed, which the receiver decides as\n \"Direct and relayed requests\" in the file header states:\n\n- Direct request. The agent's own RFC 9421 signature arrives because\n the agent sent the request itself. The verifier resolves the agent's\n keys from the COVERED `Signature-Agent` member, never from this field.\n It then MUST require this field to name that same directory: the host\n of the origin that member names, compared by\n the identity rule \"Request recipient\" defines (the shape check first,\n then case-folded, an absent port the same as \":443\", a subdomain a\n different party). A mismatch is refused as UNAUTHENTICATED with\n `request_auth_failure` SIGNATURE_INVALID: the signature verifies, but\n not for the requester the body names. A Broker applies this rule at\n BrokerService.ExecuteTransaction. At BrokerService.Resolve it refuses\n nothing for this field: every query it originates names the host of\n the origin that member names (see DiscoveryRequest.requester). An\n Exchange applies it on every request whose signer it does not admit as\n a Broker; a mismatch from a signer it does admit is a relayed request,\n one of the two cases below.\n - Purchase relayed through a Broker. BrokerService.ExecuteTransaction\n re-packages the purchase, so the request signature and the covered\n `Signature-Agent` member are the Broker's. They say only that the call comes\n from the Broker, and they sign no purchase. The Exchange MUST verify\n each item's AgentAcceptance, and the AgentRequestAcceptance, against\n the Ed25519 keys currently valid in the directory this field names\n (see AgentRequestAcceptance), never against the Broker's key. Those\n acceptances are the only agent signatures the Exchange sees, and the\n purchase belongs to the account of the directory they verify under\n (see \"Accounts\" in the file header).\n - Broker-led discovery. A ResourceQuery a Broker originated carries no agent\n signature, so nothing on that leg authenticates this field: it is the\n Broker's statement, under the Broker's own signature, of whom it\n queries for. A delegation on such a query is not honoured (see\n Delegation)."), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "id": z.string().min(1).max(255).describe("REQUIRED. A free label the agent chooses for attribution (e.g.\n \"agent-research-bot-001\"), 1 to 255 characters. Use it to tell apart the\n sub-agents or end customers that share one key directory. It is never\n identity: no verifier finds keys from it, and no party trusts it for an\n authentication or authorization decision. The requester's identity is\n `domain` together with the key that verifies. The acceptance signatures\n cover this label (see AgentAcceptancePayload), so a relaying Broker cannot\n rewrite it, but its value is still only what the agent says."), "name": z.string().describe("Human-readable name (e.g., \"Acme Research Assistant\").").optional(), "scopes": z.array(z.string()).max(64).describe("Entitlement scopes: the requester's own statement of what it is entitled\n to. On a query a Broker originated, they are the Broker's statement for the\n agent. They are signed and attributable, and the Exchange does not verify\n them. Scope coverage decides which terms the Exchange presents (see\n LicenseTerm.scopes, which also states when an Exchange MAY require a\n verified delegation); it is not access control. A false statement is\n traceable to the signing key, its directory and its account, and the\n agent's acceptance signs the scoped offer it buys; it is settled through\n reconciliation and the account relationship.\n\nEXISTENCE HIDING. A requester never learns about a resource outside its\n scopes. When the scopes leave no presentable offer for a resource, the\n Exchange answers with no offers and no absence reason (an OfferGroup with\n empty `offers` and `absence_reason` unset), exactly as for any resource\n with nothing to offer. A Broker relaying that answer adds no reason of its\n own. At purchase, an offer the Exchange presented is honoured until it\n expires, so a scope refusal cannot arise there. No reason in the protocol\n reports a scope shortfall: OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT and\n DENIAL_REASON_SCOPE_INSUFFICIENT are deprecated and never sent.\n\n Scope format: colon-separated segments, \"{domain}:{permission}\" or\n \"{profile}:{permission}\", optionally multi-segment (\"dist:US:CA\");\n matching is segment-wise per the rule below (no implicit hierarchy).\n Examples:\n \"credit:read\" — can access credit reports\n \"subscription:marketdata-2026\" — has active MarketData subscription\n \"academic:*\" — full access to academic resources\n \"internal:reports\" — can access internal reports\n \"*\" — unrestricted (public Exchange default)\n\n Matching is SEGMENT-WISE (\":\" separated). A granted scope G covers a\n required scope R iff, segment by segment, each G segment equals the\n corresponding R segment or is \"*\"; a terminal \"*\" matches all remaining\n segments, of which there must be at least one. There is NO implicit prefix\n match, and a grant NARROWER than the requirement does not cover it (G must\n be equal-to-or-broader than R), so a grant with more segments than the\n requirement never covers it, even when its last segment is \"*\". A \"*\" in R\n is a literal segment: a narrow grant never covers a wildcard requirement.\n Matching is case-sensitive and never trims. Examples: \"dist:*\" covers\n \"dist:US\", \"dist:US:CA\" and \"dist:US:*\" but not \"dist\"; \"dist:US:*\" covers\n \"dist:US:CA\" but not \"dist:EU\" or \"dist:US\"; \"publisher:*:read\" covers\n \"publisher:news:read\" but not \"publisher:news:*\"; bare \"dist\" covers only\n \"dist\"; granted \"dist:US:CA\" does NOT cover required \"dist:US\"; \"*\" covers\n everything. This same rule governs LicenseTerm.scopes and the attenuation\n of a delegation child against its parent — one algorithm protocol-wide.\n\n When empty, Exchange applies its default access policy (typically\n returns all publicly available resources).").optional(), "type": z.enum(["REQUESTER_TYPE_AGENT","REQUESTER_TYPE_HUMAN_TOOL","REQUESTER_TYPE_SERVICE","REQUESTER_TYPE_DELEGATED","REQUESTER_TYPE_RESEARCH"]).describe("What kind of entity is making this request.") }).describe("Requester identity — forwarded for authorization and audit. REQUIRED: a\n purchase names its requester, whose Requester.id and Requester.domain are\n required in turn, and every acceptance binds that requester."), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("TransactionRequest — Commit to one or more offers.\n\nAfter selecting offers, the caller commits by sending this to the\n Exchange. Supports both single-offer and batch (multi-offer) modes.\n The Exchange validates eligibility, authorizes billing, creates\n delivery, and logs each transaction.")))();
+export const TransactionRequestSchema = /* @__PURE__ */ (() => wire(z.object({ "agent_request_acceptance": z.object({ "payload": z.object({ "idempotency_key": z.string().default(""), "items": z.array(z.object({ "exchange": z.string().min(1), "offer_sig": z.string().min(1) }).describe("AgentRequestAcceptanceItem is the minimum reference needed to authorize an\n offer's membership, order, and fan-out destination without repeating the\n full Offer in every projected subrequest. Offer.signature transitively binds\n the full offer, including its exchange field; the explicit exchange lets a\n recipient derive which signed references must appear in its projection.")).min(1).max(256).describe("Complete original request order, before Broker fan-out. Capped at 256 —\n the same ceiling a discovery query's uris list carries, so one request\n can reference at most one offer per queried URI at the query cap. The Go\n verification helper enforces the same bound itself before doing any\n canonicalization work, because a verifier may run with wire validation\n off and the canonical rendering of an unbounded list is the expensive\n step an unauthenticated caller could otherwise buy for free.").optional(), "requester_domain": z.string().describe("Requester.domain of the request. Non-empty (see AgentRequestAcceptance).").default(""), "requester_id": z.string().describe("Requester.id of the request. Non-empty (see AgentRequestAcceptance).").default("") }).describe("The signed payload is carried because a projected subrequest does not carry\n offers addressed to other Exchanges and therefore cannot reconstruct the\n original complete set by itself."), "signature": z.string().min(1).describe("Hex-encoded detached Ed25519 signature over the canonical payload bytes."), "signature_algorithm": z.string().describe("Signature algorithm; \"EdDSA\" for Ed25519.").default("") }).describe("Optional for wire compatibility. When present, an Exchange verifies this\n before creating or serving request-level idempotency state. A Broker MUST\n forward it unchanged on every projected subrequest. Older clients that omit\n it retain per-item execution semantics but receive no request-level claim.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "idempotency_key": z.string().min(1).max(255).describe("Idempotency key (REQUIRED). The server MUST dedupe on this: a replay returns\n the original result rather than re-executing, and the same key with\n different items is refused with already_exists. The key is scoped to the\n agent, whatever the route: a purchase repeated directly or through any\n Broker the Exchange admits is the same purchase, and the Broker forwards the\n agent's key unchanged. A key another agent or tenant used is never a\n collision. The full rule is \"Idempotency\" in the file header. The\n transaction's durable identity is the Exchange-assigned transaction_id in\n the response."), "items": z.array(z.object({ "agent_acceptance": z.object({ "signature": z.string().min(1).describe("Hex-encoded detached Ed25519 signature over the canonical AgentAcceptancePayload\n bytes (see the canonical-signing definition on Offer.signature)."), "signature_algorithm": z.string().describe("Signature algorithm; \"EdDSA\" for Ed25519.").default("") }).describe("The agent's detached acceptance signature over this item's `offer`.\n Optional on the wire; the Exchange enforces presence per\n item at the service layer for relayed batches. Signed bytes = the canonical\n AgentAcceptancePayload form, with requester_* and idempotency_key\n taken from the ENCLOSING TransactionRequest and offer_sig = offer.signature.").optional(), "offer": z.object({ "attestations": z.array(z.object({ "attested_at": z.string().datetime({ offset: true }).describe("When this attestation was created. Agents use this to assess freshness\n (e.g., \"I accept attestations up to N hours old for breaking news\").").optional(), "claims": z.record(z.string(), z.any()).describe("Signed claims about the resource (max 4KB). A JSON object containing\n whatever properties the attesting party can determine about the resource.\n Recommended claim names for interoperability:\n estimated_quantity (integer): estimated consumption quantity (e.g., token count for text)\n word_count (integer): word count (estimated_quantity ~ word_count * 1.32 for text)\n language (string): ISO 639-1 language code\n iab_categories (string[]): IAB Content Taxonomy 3.1 codes\n content_hash (string): hash of content in \"method:hexdigest\" format\n hash_method (string): algorithm used for content_hash\n Vendors MAY add vendor-specific claims (e.g., brand_safety, sentiment).\n The protocol does NOT define \"quality score\" — it is inherently subjective.\n If a vendor provides a proprietary score, the vendor defines what it means\n via their WellKnownManifest ext[\"fora.attestation.claims_schema\"].").optional(), "keyid": z.string().describe("RFC 7638 JWK Thumbprint (the RFC 9421 keyid) of the verifier's\n attestation-signing key, resolved against the verifier's WBA directory\n (WBAFile.keys). Identifies which Ed25519 key signed this attestation.\n Enables key rotation: new keys are published with overlapping validity,\n new attestations use the new key's thumbprint, old attestations remain\n verifiable while the old key is still published.").default(""), "signature": z.string().describe("Ed25519 signature over JCS-canonicalized (RFC 8785) representation of\n {verifier, keyid, attested_at, uri, claims}. JCS (JSON Canonicalization\n Scheme) produces deterministic UTF-8 bytes: lexicographic key sorting,\n ECMAScript number serialization, strict string escaping, no whitespace.\n Each attestation is self-contained — new claim fields do not invalidate\n old attestations because the signature covers the specific claims instance.").default(""), "uri": z.string().describe("The resource URI this attestation covers. Must match the URI in the\n Offer or ResourceEntry this attestation is attached to.").default(""), "verifier": z.string().describe("Canonical domain of the attesting party (e.g., \"nytimes.com\" for\n self-attestation, \"doubleverify.com\" for third-party attestation).\n Used to look up the verifier's attestation-signing keys in its WBA\n directory (WBAFile.keys) at\n https://{verifier}/.well-known/http-message-signatures-directory").default("") }).describe("ResourceAttestation — Signed envelope of claims from a trusted party.\n\nA provider or third-party verification vendor (GumGum, DoubleVerify, IAS)\n attests to properties of the resource at a specific URI at a specific time.\n The signature covers all fields, proving origin and integrity of the claims.\n\n Verification levels (determined by who the verifier is):\n Level 0: No attestation present. Resource may carry identifiers\n (DOI, IPTC GUID via ResourceIdentity) but nothing is cryptographically\n verifiable. Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): verifier == provider domain. Provider signs\n own claims with their Ed25519 key. Agent can independently verify\n content_hash by re-computing it from delivered bytes. Requires the\n provider to serve deterministic content at the delivery endpoint.\n Level 2 (third-party attested): verifier == verification vendor domain.\n Vendor independently crawled the resource and attested to its properties.\n Agent trusts the attestation — does NOT re-verify the content hash\n (agent lacks the vendor's extraction algorithm). The Ed25519 signature\n proves the vendor made the attestation; trust is binary (\"do I trust\n this vendor?\").\n\n Claims are limited to 4KB. Attestations are carried in-memory in the\n Exchange catalog and in Offer responses — strict size limits protect\n against payload poisoning and ensure catalog performance at scale.\n\n Verifiers MUST publish their attestation-signing keys in their WBA directory\n (WBAFile.keys) at:\n https://{verifier-domain}/.well-known/http-message-signatures-directory\n identified by RFC 7638 thumbprint. Verifiers publish the claims-schema\n structure at WellKnownManifest.ext[\"fora.attestation.claims_schema\"].")).describe("Signed attestations about the resource at this URI.\n Attestations provide cryptographic proof of\n resource properties from trusted parties (providers or verification vendors).\n\nThree verification levels determine what is independently verifiable:\n Level 0 (no attestations): Resource may carry identifiers (DOI, IPTC GUID)\n for identification, but nothing is cryptographically verifiable.\n Only CDN delivery failure is auto-disputable.\n Level 1 (self-attested): Provider signs own claims with Ed25519 key.\n Agent can independently verify content hash and token count.\n CDN delivery failure + content hash mismatch are auto-disputable.\n Level 2 (third-party attested): Independent verification vendor crawled\n the resource and attested to its properties. Agent trusts the attestation\n (does not re-verify hash). Token count discrepancy is auto-disputable\n when corroborated by CDN response size.\n\n Multiple attestations may be present (e.g., provider self-attestation\n plus a third-party verification). Agents choose which to trust.").optional(), "data_as_of": z.string().datetime({ offset: true }).describe("When the offered data was current. For dynamic resources\n (resource_mutability = DYNAMIC), this is the snapshot timestamp.\n Enables the Broker to evaluate freshness: \"this credit report\n reflects data as of March 18\" or \"this drug database was updated today.\"\n\nNot set for STATIC resources (content doesn't change) or LIVE\n resources (content doesn't exist yet).\n\n The Broker compares this against RequestConstraints.max_data_age\n to filter stale offers. Example: agent requests max_data_age = 7 days,\n Broker drops offers where now() - data_as_of > 7 days.").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource will be delivered.").default(0), "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the Exchange that issued this offer (e.g.\n \"exchange.example\" or \"exchange.example:8081\"), in the form \"Request\n recipient\" defines in the file header. This is the execute-routing target:\n the agent, or a relaying Broker, sends the ExecuteTransaction call for this\n offer to this Exchange, and a Broker relaying a mixed batch groups the items\n by this value. Because it is an ordinary Offer field it falls inside the\n signed bytes (see `signature` below — the signature covers every field\n except `signature` / `signature_algorithm`), so an intermediary cannot\n redirect the execute call to a different Exchange without invalidating the\n offer, and it is what retires the X-FORA-Exchange-Endpoint transport header.\n It is also the audience statement of an ExecuteTransaction, which is why\n TransactionRequest carries no top-level `exchange`: on receipt, an Exchange\n MUST reject the request unless EVERY item's offer.exchange names its own\n domain. Presence is enforced because an empty value is unroutable — a\n relaying Broker has nothing to group or dial on, and the swap-protection\n above is vacuous when the signed bytes carry no recipient at all."), "expires_at": z.string().datetime({ offset: true }).describe("When this offer expires (ISO 8601). An offer for a term gated behind a\n verified delegation MUST NOT expire later than that delegation chain's\n effective expiry, the earliest `exp` in the chain (see Delegation).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "iab_categories": z.array(z.string()).describe("IAB Content Taxonomy category codes.\n Enables agents to filter offers by topic (e.g., \"only finance resources\").\n Uses IAB Content Taxonomy 3.1 codes.").optional(), "identity": z.object({ "c2pa_manifest": z.string().describe("C2PA content credentials manifest URI.\n Points to a sidecar or embedded C2PA manifest for this resource.\n C2PA-aware agents MAY follow this URI to validate the full provenance\n chain (creator identity, transformation history, ingredient composition)\n using C2PA libraries (JUMBF/COSE Sign1). C2PA-unaware agents can rely\n on c2pa_status and c2pa-bridged attestation claims instead.\n\nFormats:\n Sidecar: HTTPS URI to a .c2pa manifest file\n Embedded: same URI as canonical_url (manifest is inside the asset)\n Content Credentials Cloud: https://contentcredentials.org/verify?uri=...").optional(), "c2pa_status": z.enum(["C2PA_STATUS_TRUSTED","C2PA_STATUS_VALID","C2PA_STATUS_INVALID","C2PA_STATUS_ABSENT"]).describe("The full C2PA validation details (signer identity, trust list,\n action history, training/mining status) are carried in a\n ResourceAttestation with c2pa.* claims — see fora-c2pa-v1 profile.").optional(), "canonical_url": z.string().describe("Provider's authoritative URL for this resource (rel=\"canonical\").\n Always available. Different per provider for syndicated content.").optional(), "content_hash": z.string().describe("Hash of the content. Interpretation depends on hash_method:\n \"simhash-v1\" → locality-sensitive hash, for fuzzy dedup (Level 1)\n \"sha256\" → exact-match integrity hash (Level 2)\n\nLevel 1 (SimHash): computed by Exchange from extracted text.\n Agent verifies that fetched content is \"substantially similar.\"\n Tolerates dynamic page elements.\n\n Level 2 (SHA-256): computed by provider from deterministic payload.\n Agent verifies exact match. Requires provider to serve consistent\n content (e.g., API endpoint, static HTML, structured JSON).\n Mismatch = dispute. Commands premium pricing.").optional(), "doi": z.string().describe("Digital Object Identifier — persistent, never changes.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "hash_method": z.string().describe("Hash algorithm and verification level.\n Examples: \"simhash-v1\", \"minhash-v1\", \"sha256\", \"sha384\"").optional(), "iptc_guid": z.string().describe("IPTC NewsML-G2 globally unique identifier.\n Present when resource flows through news wire syndication (AP, Reuters).").optional(), "isni": z.string().describe("International Standard Name Identifier for the creator.").optional(), "resource_mutability": z.enum(["RESOURCE_MUTABILITY_STATIC","RESOURCE_MUTABILITY_DYNAMIC","RESOURCE_MUTABILITY_LIVE"]).describe("Drives hash verification behavior:\n STATIC: content_hash is stable. Agent SHOULD verify delivered content matches.\n DYNAMIC: content changes between offer and fetch (credit reports, drug databases).\n content_hash reflects state at offer generation time. Hash mismatch is\n expected and MUST NOT trigger automatic dispute.\n LIVE: content does not exist at offer time (streaming feeds, live broadcasts).\n content_hash is not applicable. The \"resource\" is the stream endpoint.\n\n Validated across 18 use cases: static content (articles, patents, legislation),\n dynamic data (credit reports, drug interactions, stock snapshots), and live\n streams (MarketData quotes, NPR broadcast, news monitoring feeds)."), "soft_binding": z.string().describe("Soft binding hash — content-derived identifier that survives format\n transcoding (resolution changes, compression, PDF-to-text extraction).\n Extracted from C2PA soft binding assertion when present.\n Enables post-delivery verification when the hard binding hash breaks\n due to legitimate format conversion.\n\nAlgorithm specified in soft_binding_method. Values are algorithm-specific\n (e.g., perceptual hash hex string, watermark identifier).").optional(), "soft_binding_method": z.string().describe("Algorithm used for soft_binding.\n Examples: \"phash-v1\" (perceptual hash), \"c2pa-watermark\" (C2PA invisible\n watermark), \"chromaprint\" (audio fingerprint).").optional() }).describe("Resource identity for cross-exchange deduplication.\n Enables Brokers to recognize the same resource offered by\n different Exchanges and compare pricing.").optional(), "offer_id": z.string().describe("Unique identifier for this offer, assigned by the Exchange.\n Opaque to the caller: not derived from the resource, its URL, or any\n other field, and carries no meaning beyond identifying this offer.\n Two offers for the same resource have different offer_ids.").default(""), "previews": z.array(z.object({ "duration": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Duration in seconds (for audio and video clips).").optional(), "height": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Height in pixels (images and video)").optional(), "media_type": z.string().describe("MIME type of the preview.\n Examples: \"image/jpeg\", \"image/webp\", \"audio/mpeg\", \"video/mp4\",\n \"text/plain\", \"application/json\"").default(""), "size": z.string().describe("Size category hint. Agents use this to select the right preview\n without fetching all of them.\n Standard values:\n \"thumbnail\" — smallest useful preview (100–150px or 5–10s)\n \"preview\" — mid-size for evaluation (300–500px or 15–30s)\n \"sample\" — larger / more detailed (for data: 1–3 sample records)").optional(), "url": z.string().describe("URL to a preview asset (thumbnail, clip, snippet, sample).\n Served by the provider's CDN, not by the Exchange.").default(""), "width": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Dimensions in pixels (for images and video).").optional() }).describe("Preview — Lightweight resource preview for offer evaluation.\n\nThe Exchange holds URLs (50–200 bytes per preview); the provider's\n CDN serves the actual bytes. This follows the universal pattern:\n Shutterstock (multi-size thumbnail URLs), Spotify (preview_url to\n 30s clip), IIIF (parameterized image URLs), OpenRTB (img.url + dims).\n\n Previews are free to fetch — no FORA transaction required. They are\n the equivalent of looking at a book cover before buying. Providers\n MAY watermark visual previews or truncate text/audio previews.\n\n The Exchange populates preview URLs during catalog ingestion. Preview\n URLs MAY be signed with a short TTL to prevent hotlinking, or public\n (provider's choice). Agents fetch previews only when evaluating\n offers, not on every discovery query.")).describe("Lightweight previews for offer evaluation.\n The Exchange holds URLs (50–200 bytes each); the provider's CDN serves\n the actual bytes. Agents fetch previews only when evaluating offers —\n not on every discovery query. Multiple previews at different sizes\n allow agents to pick the cheapest fetch for their evaluation needs.\n\nPer content type:\n Image: watermarked thumbnail (150–450px JPEG)\n Video: short clip (10–30s MP4, watermarked)\n Audio: short clip (15–30s MP3, low-bitrate or watermarked)\n Text: snippet or abstract (first 200 words as text/plain)\n Data: sample records (1–3 rows as application/json)\n Stream: optional frame capture or none (streams are priced by time)\n\n Modeled after Shutterstock (multi-size thumbnail URLs),\n Spotify (preview_url to 30s clip), IIIF (parameterized image URLs),\n and OpenRTB native (img.url + dimensions).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("The offer's price, and its ONLY price. An offer represents a single\n licensing arrangement: each projected LicenseTerm yields its own offer, and\n the Exchange moves that term's price here, so the term inside `terms`\n carries no pricing of its own (the offer.terms.pricing_unset rule). This is\n the price execute charges, the price an agent compares across Exchanges,\n and the price a metered purchase is charged at. A resource with\n multiple alternative terms (e.g. dual-licensed) produces multiple separate\n offers, one per term — never one offer with a \"headline\" picked among them.\n\nIt derives from exactly one catalog term: the term's pricing as the\n publisher declared it (ResourceEntry.terms). Each offer derives from one\n term, and the offer carries no second copy, so the offer's price and its\n term's price cannot disagree.\n\n On a metered (PER_UNIT) offer, the rate and the estimate a metered\n purchase is charged at are read from here (see Pricing). The estimate is\n optional. When the publisher states one — on the term's own pricing, or\n once for the resource on ResourceEntry.estimated_quantity — the Exchange\n carries it here as estimated_quantity, the term's own estimate taking\n precedence. When the publisher states none, the offer may carry none, and\n the purchase then charges one unit.").optional(), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Post-usage reporting requirements for this offer.\n\nWhen the term the offer sells carries an OBLIGATION_KIND_REPORTING\n obligation, the Exchange sets this from it: `required` is true (the\n offer.reporting.required_by_term rule), `window` and `required_fields` are\n the obligation's where it states them and the Exchange's own where it does\n not, and `endpoint` is the Exchange's. Otherwise the Exchange applies its\n own reporting policy.").optional(), "signature": z.string().describe("REQUIRED. Hex-encoded detached Ed25519 signature over the canonical\n serialization of the ENTIRE Offer — every field, including `pricing`,\n `terms` (the full licensing payload), `expires_at`, and `exchange`. Only\n `signature` and `signature_algorithm` are excluded from the signed bytes.\n `expires_at` is signed so the offer's validity window is\n integrity-protected: a relaying Broker cannot extend (or shorten) the TTL\n of a signed offer to replay it outside the window the Exchange intended.\n\nCANONICAL SIGNING (RFC 8785 JCS over canonical proto-JSON). The signed bytes\n are:\n\n signed_payload = JCS( protojson(msg with signature +\n signature_algorithm cleared) )\n\n i.e. render the message to canonical proto-JSON with the PINNED option set\n below, then apply RFC 8785 (JSON Canonicalization Scheme). Deterministic\n protobuf BINARY marshaling is explicitly NOT canonical across languages and\n versions (protobuf's own caveat), so it cannot be a cross-language signing\n primitive; JCS over proto-JSON can be reproduced by ANY language (Go, TS,\n Python) without a protobuf binary codec, so a broker/exchange/client in any\n language signs and verifies byte-identically. This same definition applies to\n the agent offer-acceptance signature (AgentAcceptance.signature).\n\n PINNED proto-JSON option set (the arbiter is the Go-emitted golden vector —\n whatever these options render MUST be byte-identical across all languages):\n - enum values as NAME strings (not numbers);\n - int64 / uint64 / fixed64 as decimal STRINGS;\n - bytes as standard (padded) base64;\n - google.protobuf.Timestamp / Duration per the proto-JSON WKT rules\n (RFC 3339 string for Timestamp);\n - unpopulated fields are OMITTED (never emitted as defaults);\n - field naming is snake_case (the proto field name, UseProtoNames=true),\n the naming every SDK target shares — wire, corpus, and signed form are all\n snake_case;\n - google.protobuf.Struct (`ext`) → a plain JSON object; JCS then sorts its\n keys recursively, so the Struct case needs no special handling.\n\n UNKNOWN FIELDS. A canonicalizer either OMITS content it has no schema for or\n PRESERVES it, and the rule follows from which:\n\n - OMITTING (e.g. proto-JSON, which emits only schema-defined fields): such a\n canonicalizer CANNOT reproduce the signed bytes of a message carrying\n unknown fields — what it renders silently drops part of what the signer\n covered. It MUST refuse the message rather than emit the reduced bytes,\n and a verifier built on it MUST reject rather than verify over them. The\n refusal binds at EVERY depth: a nested message and each element of a\n repeated or map field carries its own unknown-field set.\n - PRESERVING (a canonicalizer that carries unrecognized members through):\n it reproduces the signed bytes faithfully, so there is nothing to refuse.\n\n Either way an APPENDED field cannot pass: an omitting canonicalizer refuses\n the message, and a preserving one renders the appended member into bytes the\n signer never covered, so the signature fails. Without the refusal the omitting\n case would fail OPEN — an intermediary could add unknown fields to an\n already-signed message and leave its signature verifying, smuggling\n unauthenticated content through a message the recipient treats as verified.\n\n Extensions therefore ride in `ext` / `ext_critical`, which are defined fields\n and inside the signed bytes — never as undeclared field numbers.\n\n Because the signature covers `terms`, `pricing`, `expires_at`, and\n `exchange`, an intermediary (Broker) cannot tamper with price, restrictions,\n quotas, obligations, the expiry, the execute-routing target, or any\n licensing term without invalidating it.\n Agent SHOULD verify the signature (RFC 2119) against the Exchange's public\n key, and MUST reject an offer whose `expires_at` is in the past.").default(""), "signature_algorithm": z.string().describe("JOSE/JWA algorithm identifier (RFC 8037 §3.1). Always 'EdDSA' for\n Ed25519. Advisory only: this field is cleared before the canonical\n payload is signed, so it is not covered by the signature.").default(""), "subscription_id": z.string().describe("If set, this offer is available under an existing subscription/deal.\n No per-request billing — usage tracked against subscription quota.\n Pricing.rate = \"0\" for subscription offers (zero marginal cost).\n The Broker SHOULD prefer subscription offers when available.").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Subscription quota state, when this offer is under a subscription.\n Enables the agent to see remaining quota before committing.\n Multiple entries when the subscription has independent quotas\n (e.g., access count + spend cap).").optional(), "terms": z.array(z.object({ "license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("Governing license document. Authoritative for REFERENCE_ONLY terms, which\n MUST carry a License with a non-empty uri — a REFERENCE_ONLY term that\n references nothing is rejected at ingest.").optional(), "obligations": z.array(z.object({ "detail": z.string().describe("Free-form detail: attribution string, notice file URI, etc.\n OBLIGATION_KIND_OTHER without it → lint warning.").optional(), "kind": z.enum(["OBLIGATION_KIND_ATTRIBUTION","OBLIGATION_KIND_CONTRIBUTION","OBLIGATION_KIND_SHARE_ALIKE","OBLIGATION_KIND_NETWORK_COPYLEFT","OBLIGATION_KIND_NOTICE","OBLIGATION_KIND_OTHER","OBLIGATION_KIND_REPORTING"]).describe("What the agent must do."), "reporting": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("The reporting requirement of an OBLIGATION_KIND_REPORTING obligation: the\n window within which the agent files ReportUsage and the fields the report\n must carry. Set only on a REPORTING obligation (the\n obligation.reporting.requires_reporting_kind rule). `endpoint` is not set\n here (obligation.reporting.no_endpoint) and `required` is not read: a\n REPORTING obligation always requires a report. Absent, the obligation still\n requires a report, under the Exchange's own window and fields. The Exchange\n carries the obligation onto Offer.reporting (see that field).").optional(), "scope_license": z.object({ "id": z.string().describe("Stable short identifier: SPDX short-id (\"GPL-3.0-only\"), TollBit cuid,\n or catalog doc-id. Used by agents and the vocab linter for known-license\n lookup; SHARE_ALIKE derivatives default their scope_license to this.").optional(), "immutable": z.boolean().describe("Data-labels TDL: the document at uri is versioned and will not change.").optional(), "name": z.string().describe("Human-readable name (licenseType, schema.org node name).").optional(), "uri": z.string().describe("Canonical identity of the license document (RFC 3986). MUST NOT be\n URL-validated — data-labels TDL identifiers use non-URL schemes.\n For REFERENCE_ONLY terms this is the authoritative specification.\n Examples:\n \"https://creativecommons.org/licenses/by/4.0/\"\n \"https://news.example/licensing/ai-terms-2026\"\n\n\"MUST NOT URL-validate\" means do not REJECT non-URL schemes — it does NOT\n mean fetch blindly. A consumer that dereferences this URI MUST apply these\n SSRF guards: scheme allowlist, block loopback/private/metadata addresses\n (resolve-then-check), fetch via an egress proxy, and treat the response\n as untrusted content. Verify the fetched bytes against `uri_digest`\n before use.").optional(), "uri_digest": z.string().regex(new RegExp("^(sha256:[0-9a-f]{64}|sha384:[0-9a-f]{96}|sha512:[0-9a-f]{128})?$")).describe("Cryptographic digest of the document at `uri`, in \"method:hexdigest\" form\n (e.g. \"sha256:9f86d081...\"). Pins the referenced document so a consumer can\n verify the bytes it fetches match what was offered; covered by the offer\n signature, so it is tamper-evident end to end. REQUIRED whenever `uri` is\n non-empty — any semantics, mutable or not: without a pinned digest a MitM\n (or the publisher) can swap the document the agent reads. The party that\n sets uri supplies uri_digest with it: the publisher on a term it pushes.\n\nThe method MUST be a collision-resistant hash — sha256, sha384, or sha512.\n Legacy md5/sha1 are rejected on the wire: a forgeable digest would defeat\n the swap-protection this field exists for. The CEL is STRUCTURE ONLY\n (allowlisted prefix + matching hex length); presence (digest-when-uri) is\n enforced at ingest.").optional() }).describe("The license that derivatives must be released under. REQUIRED for\n SHARE_ALIKE (rejected if absent), where it MUST identify a license — set\n `id` (SPDX short-id, the common copyleft case, often the term's own\n License.id) and/or `uri`. Because it is a License, a referenced `uri`\n inherits the uri_digest swap-protection rule: a uri without a digest is\n rejected, exactly as for any other license reference.").optional(), "trigger": z.enum(["OBLIGATION_TRIGGER_ON_USE","OBLIGATION_TRIGGER_ON_DISTRIBUTION","OBLIGATION_TRIGGER_ON_NETWORK_SERVICE","OBLIGATION_TRIGGER_ON_DERIVATIVE"]).describe("When the obligation activates.") }).describe("Obligation — A post-use behavioral requirement attached to a LicenseTerm.\n\nExamples:\n Attribution on display: cite the author whenever content is shown to a user.\n Share-alike on derivative: AI-generated content that incorporates this work\n must be released under the same license.\n Notice on distribution: include the copyright notice when distributing copies.")).max(64).describe("Post-use behavioral requirements.\n At most 64, for the reason quotas carries.").optional(), "part_label": z.string().describe("Informational human-readable name for this sub-part (sub-part terms).").optional(), "pricing": z.object({ "currency": z.string().describe("ISO 4217 currency code (e.g. \"USD\", \"EUR\").").default(""), "estimated_quantity": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Estimated quantity in the metering unit (`unit`).\n For text: token count. For video: duration in seconds.\n For documents: page count. For data: record count.\n\nOPTIONAL everywhere: on a LicenseTerm a publisher pushes and on an Offer.\n It is E in the metered price rule above. When a metered offer states it,\n it is positive (the offer.metered.estimate_positive rule), and the\n purchase charges E × rate. When a metered offer states none, the purchase\n charges one unit, 1 × rate. A PER_UNIT price may state a rate\n with or without an estimate; whether to state one is the publisher's\n decision, and nothing requires it.\n\n The estimate describes the resource rather than the arrangement, so a\n publisher may state it once on ResourceEntry.estimated_quantity instead of\n on each term. An estimate the publisher states either way is carried onto\n the offer's pricing (see Offer.pricing).").optional(), "license_duration_months": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("License duration in months. How long the granted access remains valid.").optional(), "metering": z.enum(["PRICING_METERING_ONLINE","PRICING_METERING_NONE","PRICING_METERING_OFFLINE_SELF_REPORTED"]).describe("How usage is tracked for billing reconciliation.\n Absent = PRICING_METERING_ONLINE (default real-time tracking).\n NONE = one-time perpetual sale; no ReportUsage required after ExecuteTransaction.\n OFFLINE_SELF_REPORTED = agent self-reports physical-world consumption.").optional(), "model": z.enum(["PRICING_MODEL_FREE","PRICING_MODEL_PER_UNIT","PRICING_MODEL_FLAT"]).describe("Provider's pricing model."), "rate": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Price in the provider's model, as an exact decimal string — e.g. \"0.05\" =\n $0.05 per article. NOT a float: money is decimal to avoid binary rounding and\n to allow arbitrary sub-cent precision (e.g. \"0.0001234\"). Denominated in `currency`.").default(""), "unit": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)?$")).max(64).describe("Metering basis — the \"per what\" of PER_UNIT pricing. REQUIRED when\n model = PER_UNIT. Custom units namespace as \"vendor:unit\". Ignored for\n FREE / FLAT.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare tokens. A buf plugin reads them structurally and emits the\n pricingunits constants + IsRegistered; ingest enforces membership from\n those. The CEL is STRUCTURE ONLY (empty / bare-form / vendor:namespaced) —\n it never lists the tokens, so it cannot drift from the registry.").optional(), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Normalized cost per unit — the universal comparison metric, exact decimal string.\n For text: cost per token. For video: cost per second.\n For data: cost per record. For APIs: cost per call.\n Denominated in the Exchange's base_currency (from its WellKnownManifest).").optional() }).describe("Pricing for this term. Where it is required depends on the message that\n holds the term, so the requirement is a rule of that message rather than of\n this field:\n - On a catalog term (ResourceEntry.terms) it is REQUIRED regardless of\n semantics — an agent cannot act on a priceless term, so absent Pricing\n is a validation error at ingest (resource_entry.terms.pricing_required).\n model = FREE must be stated explicitly (absent Pricing is not free). A\n REFERENCE_ONLY term states its price here too; its License governs the\n human-readable terms but does not replace the machine-readable price.\n - On the term an offer carries (Offer.terms) it MUST be unset\n (offer.terms.pricing_unset): the offer's price is Offer.pricing, stated\n once.").optional(), "quotas": z.array(z.object({ "limit": z.coerce.number().int().gte(1).describe("Maximum allowed value in the given window. A quota of 0 grants\n nothing — express \"no access\" by omitting the term, not a zero quota."), "metric": z.string().regex(new RegExp("^([a-z0-9-]+|[A-Za-z0-9._-]+:[A-Za-z0-9._-]+)$")).max(64).describe("The unit being capped — an open vocabulary axis.\n\nThe (fora.v1.vocab) entries below are the SOLE authored source of the\n registered bare metric tokens. A buf plugin reads them structurally and\n emits the quotametrics constants + IsRegistered; ingest enforces membership\n from those. The CEL is STRUCTURE ONLY (non-empty bare token or\n vendor:namespaced) — it never lists the tokens, so it cannot drift.\n\n Token meanings:\n display-words Words of content text rendered to an end user.\n impressions Times the content is displayed to an end user.\n tokens LLM output tokens generated using this content.\n input-tokens LLM input tokens consumed from this content.\n units-manufactured Physical units manufactured from this design/pattern.\n accesses Distinct content access / retrieval events.\n copies Digital or physical copies produced.\n seats Distinct named users licensed to access the content."), "window": z.enum(["QUOTA_WINDOW_HOURLY","QUOTA_WINDOW_DAILY","QUOTA_WINDOW_MONTHLY","QUOTA_WINDOW_TOTAL"]).describe("Time window over which the limit accumulates.") }).describe("Quota — A usage cap the term declares.\n\nA quota states how much a licensee may consume under the term before the\n term expires or must be renegotiated. Like a restriction, it is a declared\n term the parties agree to. Quotas are NOT billing quantities — billing is in\n Pricing.\n\n The metric vocabulary is authored ONLY in the (fora.v1.vocab) entries on\n Quota.metric below; the quotametrics constants + IsRegistered derive from it.")).max(64).describe("Usage caps. The agent must not exceed any individual Quota.\n At most 64, the bound every per-message list in this contract carries when\n no rule walks it more than once. It bounds what one term may carry, not the\n work of checking one — a validator walks every element it is handed before\n the cap is reported, so the cost of checking is bounded at the transport.").optional(), "restrictions": z.array(z.object({ "advisory": z.boolean().describe("Fail-closed by default. When false (the default), this restriction is\n BINDING: an agent that cannot evaluate every token in it — including an\n unknown vendor token — MUST decline the term. Set advisory = true to\n downgrade an unverifiable restriction to non-blocking. This deliberately\n inverts the COSE-`crit` opt-in default: a license restriction a consumer\n does not understand should stop it, not be silently ignored.").default(false), "kind": z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"]).describe("Which dimension this restriction applies to. Defined-only: the axis set is\n CLOSED, and a number outside it is refused rather than ignored. A custom\n axis is RESTRICTION_KIND_OTHER, whose meaning rides in permitted/prohibited,\n so a new number was never the extension mechanism — accepting one would\n admit a restriction no consumer can evaluate onto a term whose default is\n BINDING (see advisory below), which fails open on the axis a publisher most\n needs honoured. Closing the axis does NOT bound the cost of the one-per-kind\n rule below, and must not be read as doing so: a number this rule refuses is\n still distinct from every other, so that rule's all() finds no duplicate to\n stop on and walks the list in full anyway. Its cost is bounded by the size\n test the rule itself carries."), "permitted": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens allowed on this axis. Empty = all permitted.\n For FUNCTION: \"ai-input\", \"ai-train\", \"search\", \"editorial\", \"commercial\", …\n For GEOGRAPHY: \"US\", \"DE\", \"EU\", \"EEA\", \"*\", …\n For USER_TYPE: \"individual\", \"academic\", \"commercial_entity\", …").optional(), "prohibited": z.array(z.string().regex(new RegExp("^[A-Za-z0-9._:*-]+$")).min(1).max(64)).max(64).describe("Tokens blocked on this axis. Takes precedence over permitted[].").optional() }).describe("Restriction — A single constraint on one licensing dimension.\n\nRestrictions model allowed and prohibited values on one axis (function,\n geography, or user-type). They are validated and normalized at ingest and\n RIDE ON THE OFFER: they are declared terms the parties agree to when the\n agent accepts the offer. The AGENT self-selects the term whose restrictions it\n can honour and bears compliance. The Exchange never enforces a restriction:\n it is NOT an Exchange-side gate the requester must pass to see or buy a term.\n\n An Exchange or Broker MAY, purely as a CONVENIENCE, pre-filter the offers it\n returns against the limits the query states in ResourceQuery.acceptable_restrictions\n (the same RestrictionKind axes/vocabulary the terms use) — e.g. an agent that\n only wants US-eligible content can ask the Exchange to skip the rest so it\n doesn't pay to discover offers it would never accept. That filter is advisory and\n optional: a different Broker may not apply it, and it is a recommendation\n matched to the request, never an enforcement verdict. When an Exchange does\n drop offers this way it MAY signal it via OfferAbsenceReason.RESTRICTION_FILTERED\n (with the axes in OfferGroup.restriction_filters). Term visibility is otherwise\n gated only by resource_id/URI and by the scopes in the requester's signed\n request — see LicenseTerm.scopes.\n\n Reading a restriction:\n A value is in-scope when it matches at least one permitted[] token\n AND matches none of the prohibited[] tokens.\n Empty permitted[] = any value is permitted on this axis.\n Empty prohibited[] = nothing is explicitly prohibited.\n\n Vocabulary sources (authored on the RestrictionKind enum values via\n (fora.v1.vocab_enum); the functiontokens / geographytokens / usertypes\n constants + IsRegistered derive from them):\n FUNCTION — RSL 1.0 AI-use vocabulary + established IP/copyright terms\n GEOGRAPHY — ISO 3166-1 alpha-2 (structural) + the specials *, EU, EEA\n USER_TYPE — FORA user/organization categories")).max(8).describe("Usage restrictions (function, geography, user-type).\n Multiple restrictions are AND-combined — the agent must satisfy all of them.\n At most 8, and this list is the one of the three that does NOT carry the\n contract's usual 64: only one restriction per axis is valid, the axis enum\n is defined-only, so four is the longest conformant list and eight leaves\n room for an axis this version does not have. Like the caps on quotas and\n obligations, this one bounds the DOCUMENT — how many restrictions one term\n may carry — and not the work of checking it: a validator walks every element\n it is handed before any cardinality rule is reported, so an over-cap list is\n traversed in full on its way to being refused.\n\nWhat makes this list different is that one rule walks it against ITSELF. The\n one-per-kind rule below is quadratic, so it carries its own size test and\n stays silent above this cap; a conformance guard holds the two numbers equal,\n because a cap raised without the test would leave the lists in between\n unchecked for duplicate axes and accepted. The neighbouring disjointness rule\n on each element is quadratic only in that element's two token lists, both\n capped at 64, so its cost is bounded per restriction and linear across the\n list — it needs no such test.").optional(), "scopes": z.array(z.string()).max(64).describe("Scope-gating: the Exchange presents an offer for this term only when the\n scopes in the requester's signed request (Requester.scopes) cover ALL of\n these scopes (AND-semantics). Empty = public. A subscription term is\n Pricing{model:FREE} + scopes:[\"subscription:...\"].\n\nRequester.scopes is the requester's own statement, which the Exchange does\n not verify, so a scope gate hides a term from requesters that do not state\n the scope; it does not stop one that states it falsely. An Exchange MAY\n instead gate a term behind a verified delegation (a subscription term is\n the usual case): then only the delegation chain's verified effective scopes\n are matched against these scopes, and Requester.scopes is ignored for that\n term. To a requester that presents no valid delegation, it answers as if\n the term did not exist, as existence hiding requires (see\n Requester.scopes). An offer the Exchange presented is honoured until it\n expires, and a gated offer expires no later than the chain's effective\n expiry (see Delegation).\n\n Coverage uses the SAME matching rule as Requester.scopes:\n segment-wise (\":\" separated), each granted segment must equal the\n corresponding required segment or be \"*\", a terminal \"*\" matches all\n remaining segments (at least one), and there is NO implicit prefix match\n (a grant narrower than the requirement does not cover it). \"dist:*\" covers\n \"dist:US\" and \"dist:US:CA\"; \"dist\" covers only \"dist\". There is exactly\n one scope-matching algorithm across the protocol.").optional(), "semantics": z.enum(["TERM_SEMANTICS_ENUMERATED","TERM_SEMANTICS_REFERENCE_ONLY"]).describe("How to interpret the machine fields.") }).describe("LicenseTerm — Universal licensing unit.\n\nOne LicenseTerm describes one complete access arrangement for a resource.\n A resource carries zero or more terms; having multiple terms is the normal\n case (one per use category, user type, or commercial arrangement).\n\n The same LicenseTerm shape appears at ingestion (ResourceEntry.terms) and\n at emission (Offer.terms). The Exchange stores what the publisher pushed,\n with each restriction token in its canonical form (trimmed, case-folded and\n alias-resolved, as the ingest tier below describes), and surfaces it on\n discovery, so agents see the terms the publisher declared, with no other\n translation or reformulation — and one move: on an offer the term's price is\n Offer.pricing, and the term itself carries none.\n\n Validation rules:\n - Pricing MUST be present on EVERY catalog term (ResourceEntry.terms),\n regardless of semantics — the resource_entry.terms.pricing_required\n rule. Absent Pricing → reject at ingest: an agent cannot act on a term\n with no price. This holds for REFERENCE_ONLY too — its License governs\n the human-readable terms, but the machine-readable price is still stated\n here, not deferred to the document.\n - Pricing MUST be unset on the term an offer carries (Offer.terms) — the\n offer.terms.pricing_unset rule. The offer's price is Offer.pricing, and\n each offer derives from exactly one catalog term, so the price is stated\n once and cannot disagree with itself. The requirement is therefore a\n rule of the message that holds the term, not of LicenseTerm.\n - model=FREE must be explicit. Absent Pricing ≠ free. A term may be FREE\n under an arbitrary license; the agent still needs the price stated so it\n knows the access is free rather than unpriced.\n - A PER_UNIT term MAY state Pricing.estimated_quantity. It is not\n required, on the pushed term or on the offer it projects to — see\n Pricing.estimated_quantity.\n - REFERENCE_ONLY terms MUST carry a License with a non-empty uri. A\n REFERENCE_ONLY term that references no document is meaningless → reject\n at ingest.\n - Restriction tokens are validated against the vocab registry.\n Unknown tokens produce a PushResourcesResponse.warnings[] entry\n but do NOT cause rejection (forward-compatible).")).min(1).max(1).describe("The licensing term this offer sells, sourced from the publisher's\n ResourceEntry. EXACTLY ONE, and the bound is enforced rather than asked\n for: an offer IS a single licensing arrangement, so N terms on a resource\n project to N offers on that resource, each selling one of them — see\n `pricing` above. An offer with no term would be an offer projected from\n nothing, which is why the floor is one and not zero. The field is\n `repeated` because it shipped that way in v1 and renumbering a released\n field is not available to us, so the cardinality rides as a validation\n rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a\n resource carries many terms, an offer sells one of them.\n\nThe distinction is not cosmetic. `pricing` is what an agent sees and what\n execute charges, so a term that is not on its own offer has no price an\n agent can see and no offer_id an agent can buy. Fusing several\n terms onto one offer makes every term but the first unsellable — a\n dual-licensed resource sells only under whichever arrangement the\n publisher happened to store first.\n\n The term here carries everything the publisher declared EXCEPT its price:\n its `pricing` MUST be unset (the offer.terms.pricing_unset rule), because\n the offer's price is `pricing` above. Its semantics, license,\n restrictions, quotas, obligations, scopes and part_label are the\n publisher's, unchanged.\n\n Where a term is reachable only under an existing subscription, its offer\n carries `subscription_id` and prices at zero marginal cost; the terms a\n requester may see at all are selected by `LicenseTerm.scopes`.\n\n See: Universal Licensing Core section.").optional(), "title": z.string().describe("Resource title (human-readable, for display/logging).").optional() }).describe("The FULL signed Offer for this batch entry, reflected back exactly as\n received at discovery. The Exchange verifies `offer.signature` over these\n presented bytes — stateless, no reconstruct-from-catalog. REQUIRED: every\n batch item carries its offer.") }).describe("TransactionItem — A single offer commitment within a batch transaction.")).min(1).describe("The offers committed in this request (REQUIRED, min 1), each carrying its\n own reflected signed Offer + detached acceptance. A single offer is the\n degenerate 1-element list. The Exchange verifies each item's\n `offer.signature` (which covers pricing, terms, and expires_at) over the\n presented bytes against its own key — stateless, self-contained bearer\n tokens, with no reconstruct-from-catalog.").optional(), "requester": z.object({ "delegation": z.object({ "expires_at": z.string().datetime({ offset: true }).describe("Mirror of the chain's earliest `exp`. The Exchange MUST reject an expired\n token; the signed `exp` values decide, not this field.").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "issuer": z.string().describe("Mirror of the ROOT token's `iss`, the identity configured with the trusted\n key (never an intermediate or terminal issuer). Informational: the\n Exchange never discovers a key from it, through OIDC, JWKS or otherwise.").optional(), "max_accesses": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Maximum number of accesses allowed under this delegation.\n Exchange tracks cumulative access count against this cap.\n Deny with DENIAL_REASON_QUOTA_EXCEEDED when count >= limit.\n For subscriptions with \"10,000 accesses/month\", this carries the ceiling.").optional(), "max_spend_cents": z.coerce.number().int().describe("Maximum spend in currency minor units (e.g., cents for USD).\n Exchange tracks cumulative spend against this cap.").optional(), "principal_domain": z.string().describe("Mirror of the ROOT token's `iss`, the configured trust anchor's identity.\n Informational: no key is looked up from it, and it establishes no trust.").default(""), "principal_id": z.string().describe("Principal's identifier (e.g., \"user@acme.com\", \"marketdata.example.com\").\n Informational; not read by the verifier.").default(""), "quota_period": z.string().describe("Quota reset period. How often the access/spend counters reset.\n Example: 30 days for monthly subscriptions — \"2592000s\" on the wire\n (proto-JSON encodes Duration as seconds; \"720h\" is not accepted).\n When absent, the quota is lifetime (bounded only by expires_at).").optional(), "revocation_uri": z.string().describe("Informational. Revocation lookup is not part of the delegation profile:\n the verifier never fetches this URI. A chain is invalidated early by\n withdrawing the configured root key.").optional(), "scopes": z.array(z.string()).describe("Mirror of the terminal token's effective scopes. In the chain, each\n child's scopes are covered by its parent's under the scope-matching rule\n on Requester.scopes (attenuation — can only narrow, not widen).").optional(), "token": z.string().regex(new RegExp("^[A-Za-z0-9+/]{0,43691}={0,2}$")).max(43692).describe("Token bytes. A JWT (base64url-encoded JWS). At most 32768 bytes: the\n complete chain, every embedded parent included, must fit. The limit is a\n wire rule, checked before the token is parsed and independent of whether\n the bytes are a valid JWT; a longer token is an invalid request, not a\n per-item denial.").default(""), "token_format": z.string().describe("Token format: \"jwt\" (default). Empty is treated as \"jwt\". The field stays\n open for a future format.").default("") }).describe("Optional delegation — present when the requester acts on behalf of\n another entity (user, organization, upstream agent). On a query a Broker\n originated it is not honoured, because the Broker signs alone and nothing\n on that leg can pass its holder binding; the Exchange answers as if none\n were present (see Delegation).").optional(), "domain": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("REQUIRED. Bare host of the requester's key directory: the WBA directory at\n {domain}/.well-known/http-message-signatures-directory that publishes the\n agent's Ed25519 keys. It carries the bare-host shape \"Request recipient\"\n defines in the file header (a port allowed), for the same structural\n reason: a scheme, path or query smuggled in here would choose what gets\n fetched, not merely from where. It is never a free label. Every verifier\n reads it as the name of that directory, and the rule that binds it depends\n on whether the request is direct or relayed, which the receiver decides as\n \"Direct and relayed requests\" in the file header states:\n\n- Direct request. The agent's own RFC 9421 signature arrives because\n the agent sent the request itself. The verifier resolves the agent's\n keys from the COVERED `Signature-Agent` member, never from this field.\n It then MUST require this field to name that same directory: the host\n of the origin that member names, compared by\n the identity rule \"Request recipient\" defines (the shape check first,\n then case-folded, an absent port the same as \":443\", a subdomain a\n different party). A mismatch is refused as UNAUTHENTICATED with\n `request_auth_failure` SIGNATURE_INVALID: the signature verifies, but\n not for the requester the body names. A Broker applies this rule at\n BrokerService.ExecuteTransaction. At BrokerService.Resolve it refuses\n nothing for this field: every query it originates names the host of\n the origin that member names (see DiscoveryRequest.requester). An\n Exchange applies it on every request whose signer it does not admit as\n a Broker; a mismatch from a signer it does admit is a relayed request,\n one of the two cases below.\n - Purchase relayed through a Broker. BrokerService.ExecuteTransaction\n re-packages the purchase, so the request signature and the covered\n `Signature-Agent` member are the Broker's. They say only that the call comes\n from the Broker, and they sign no purchase. The Exchange MUST verify\n each item's AgentAcceptance, and the AgentRequestAcceptance, against\n the Ed25519 keys currently valid in the directory this field names\n (see AgentRequestAcceptance), never against the Broker's key. Those\n acceptances are the only agent signatures the Exchange sees, and the\n purchase belongs to the account of the directory they verify under\n (see \"Accounts\" in the file header).\n - Broker-led discovery. A ResourceQuery a Broker originated carries no agent\n signature, so nothing on that leg authenticates this field: it is the\n Broker's statement, under the Broker's own signature, of whom it\n queries for. A delegation on such a query is not honoured (see\n Delegation)."), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "id": z.string().min(1).max(255).describe("REQUIRED. A free label the agent chooses for attribution (e.g.\n \"agent-research-bot-001\"), 1 to 255 characters. Use it to tell apart the\n sub-agents or end customers that share one key directory. It is never\n identity: no verifier finds keys from it, and no party trusts it for an\n authentication or authorization decision. The requester's identity is\n `domain` together with the key that verifies. The acceptance signatures\n cover this label (see AgentAcceptancePayload), so a relaying Broker cannot\n rewrite it, but its value is still only what the agent says."), "name": z.string().describe("Human-readable name (e.g., \"Acme Research Assistant\").").optional(), "scopes": z.array(z.string()).max(64).describe("Entitlement scopes: the requester's own statement of what it is entitled\n to. On a query a Broker originated, they are the Broker's statement for the\n agent. They are signed and attributable, and the Exchange does not verify\n them. Scope coverage decides which terms the Exchange presents (see\n LicenseTerm.scopes, which also states when an Exchange MAY require a\n verified delegation); it is not access control. A false statement is\n traceable to the signing key, its directory and its account, and the\n agent's acceptance signs the scoped offer it buys; it is settled through\n reconciliation and the account relationship.\n\nEXISTENCE HIDING. A requester never learns about a resource outside its\n scopes. When the scopes leave no presentable offer for a resource, the\n Exchange answers with no offers and no absence reason (an OfferGroup with\n empty `offers` and `absence_reason` unset), exactly as for any resource\n with nothing to offer. A Broker relaying that answer adds no reason of its\n own. At purchase, an offer the Exchange presented is honoured until it\n expires, so a scope refusal cannot arise there. No reason in the protocol\n reports a scope shortfall: OFFER_ABSENCE_REASON_SCOPE_INSUFFICIENT and\n DENIAL_REASON_SCOPE_INSUFFICIENT are deprecated and never sent.\n\n Scope format: colon-separated segments, \"{domain}:{permission}\" or\n \"{profile}:{permission}\", optionally multi-segment (\"dist:US:CA\");\n matching is segment-wise per the rule below (no implicit hierarchy).\n Examples:\n \"credit:read\" — can access credit reports\n \"subscription:marketdata-2026\" — has active MarketData subscription\n \"academic:*\" — full access to academic resources\n \"internal:reports\" — can access internal reports\n \"*\" — unrestricted (public Exchange default)\n\n Matching is SEGMENT-WISE (\":\" separated). A granted scope G covers a\n required scope R iff, segment by segment, each G segment equals the\n corresponding R segment or is \"*\"; a terminal \"*\" matches all remaining\n segments, of which there must be at least one. There is NO implicit prefix\n match, and a grant NARROWER than the requirement does not cover it (G must\n be equal-to-or-broader than R), so a grant with more segments than the\n requirement never covers it, even when its last segment is \"*\". A \"*\" in R\n is a literal segment: a narrow grant never covers a wildcard requirement.\n Matching is case-sensitive and never trims. Examples: \"dist:*\" covers\n \"dist:US\", \"dist:US:CA\" and \"dist:US:*\" but not \"dist\"; \"dist:US:*\" covers\n \"dist:US:CA\" but not \"dist:EU\" or \"dist:US\"; \"publisher:*:read\" covers\n \"publisher:news:read\" but not \"publisher:news:*\"; bare \"dist\" covers only\n \"dist\"; granted \"dist:US:CA\" does NOT cover required \"dist:US\"; \"*\" covers\n everything. This same rule governs LicenseTerm.scopes and the attenuation\n of a delegation child against its parent — one algorithm protocol-wide.\n\n When empty, Exchange applies its default access policy (typically\n returns all publicly available resources).").optional(), "type": z.enum(["REQUESTER_TYPE_AGENT","REQUESTER_TYPE_HUMAN_TOOL","REQUESTER_TYPE_SERVICE","REQUESTER_TYPE_DELEGATED","REQUESTER_TYPE_RESEARCH"]).describe("What kind of entity is making this request.") }).describe("Requester identity — forwarded for authorization and audit. REQUIRED: a\n purchase names its requester, whose Requester.id and Requester.domain are\n required in turn, and every acceptance binds that requester."), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("TransactionRequest — Commit to one or more offers.\n\nAfter selecting offers, the caller commits by sending this to the\n Exchange. Supports both single-offer and batch (multi-offer) modes.\n The Exchange validates eligibility, authorizes billing, creates\n delivery, and logs each transaction.")))();
export const TransactionResponseSchema = /* @__PURE__ */ (() => wire(z.object({ "agent_identity_hash": z.string().describe("Identity that a delivered retrieval_endpoint is bound to: the RFC 7638 JWK\n Thumbprint of the agent's Ed25519 request-signing key (see \"Retrieval-URL\n identity binding\" above). On a sub-request a Broker re-packaged, the request\n signer is the Broker, and this is the thumbprint of the agent key the\n AgentAcceptances verify under instead. Shared across the request; set once.").default(""), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "items": z.array(z.object({ "billing_id": z.string().describe("Billing record identifier minted by the Exchange's billing adapter for\n this transaction (not the account handle — see RegisterResponse.billing_ref).").default(""), "cost": z.object({ "amount": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Exact decimal string (not a float), e.g. \"19.99\". Denominated in `currency`.").default(""), "currency": z.string().default(""), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).optional() }).describe("Cost for this item. For a metered (PER_UNIT) item this is the purchase\n charge: the offer's estimated_quantity × rate, or one unit, 1 × rate, when\n the offer states no estimate (see Pricing).").optional(), "delivery_method": z.union([z.string().regex(new RegExp("^DELIVERY_METHOD_UNSPECIFIED$")), z.enum(["DELIVERY_METHOD_DIRECT","DELIVERY_METHOD_INSTRUCTIONS","DELIVERY_METHOD_STREAMING"]), z.coerce.number().int().gte(-2147483648).lte(2147483647)]).describe("How resource is delivered for this item.").default(0), "denial_reason": z.enum(["DENIAL_REASON_ACCOUNT_INACTIVE","DENIAL_REASON_INSUFFICIENT_BALANCE","DENIAL_REASON_RATE_LIMITED","DENIAL_REASON_CONTENT_UNAVAILABLE","DENIAL_REASON_RESTRICTION_NOT_SATISFIED","DENIAL_REASON_REPORTING_OVERDUE","DENIAL_REASON_OFFER_EXPIRED","DENIAL_REASON_SIGNATURE_INVALID","DENIAL_REASON_QUOTA_EXCEEDED","DENIAL_REASON_DELEGATION_INVALID","DENIAL_REASON_SCOPE_INSUFFICIENT","DENIAL_REASON_ENTITLEMENT_MISSING","DENIAL_REASON_ENTITLEMENT_MALFORMED","DENIAL_REASON_ENTITLEMENT_EXPIRED","DENIAL_REASON_ENTITLEMENT_WRONG_BUYER","DENIAL_REASON_SUBSCRIPTION_LAPSED","DENIAL_REASON_ENTITLEMENT_NOT_GRANTED","DENIAL_REASON_ACCOUNT_NOT_REGISTERED","DENIAL_REASON_RELAY_NOT_ACCEPTED","DENIAL_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("Set if this specific item was denied (others may succeed). Every per-item\n denial is answered here, in a successful response, also when the request\n carried only this item: a denied one-item purchase is never a non-OK error.").optional(), "expires_at": z.string().datetime({ offset: true }).describe("When retrieval_endpoint expires.").optional(), "offer_id": z.string().describe("The offer_id this result is for.").default(""), "refusal": z.object({ "code": z.string().min(1).max(64).describe("The Connect code the party answered with, in its wire form, e.g.\n \"permission_denied\" or \"unauthenticated\". When the party gave no answer\n — unreachable, or the call timed out — it is the code the relaying party's\n call failed with (\"unavailable\", \"deadline_exceeded\"). A refusal with\n either of those codes is an ambiguous outcome: a purchase may have\n completed. On a Broker purchase, retrying the agent's request with the same\n idempotency_key, on any route, is answered from each Exchange's stored\n result, so it settles the outcome without buying twice."), "detail": z.object({ "catalog_rejection": z.object({ "reason": z.enum(["CATALOG_REJECTION_REASON_NOT_CATALOG_CONTRIBUTOR","CATALOG_REJECTION_REASON_TENANT_MISMATCH","CATALOG_REJECTION_REASON_DOMAIN_NOT_VERIFIED","CATALOG_REJECTION_REASON_SIGNATURE_INVALID","CATALOG_REJECTION_REASON_MALFORMED_ENTRY","CATALOG_REJECTION_REASON_UNKNOWN_VOCAB_TOKEN","CATALOG_REJECTION_REASON_QUOTA_EXCEEDED","CATALOG_REJECTION_REASON_TERMS_LIMIT_EXCEEDED","CATALOG_REJECTION_REASON_URI_UNAVAILABLE","CATALOG_REJECTION_REASON_UNKNOWN_CRITICAL_EXTENSION","CATALOG_REJECTION_REASON_EXCHANGE_NOT_LISTED"]).describe("The rejection reason (defined-only, non-zero)"), "rejected_paths": z.array(z.string()).describe("The entry paths the refusal is about. A catalog push is all-or-nothing, so\n these name which entries failed inside a submission that persisted nothing —\n they are not a list of what was dropped from an otherwise applied batch.").optional() }).describe("`reason` oneof — CatalogService rejection").optional(), "dispute_failure": z.object({ "reason": z.enum(["DISPUTE_FAILURE_REASON_TRANSACTION_NOT_FOUND","DISPUTE_FAILURE_REASON_REPORT_NOT_FILED","DISPUTE_FAILURE_REASON_WINDOW_EXPIRED","DISPUTE_FAILURE_REASON_DUPLICATE","DISPUTE_FAILURE_REASON_INELIGIBLE","DISPUTE_FAILURE_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("The failure reason (defined-only, non-zero)") }).describe("`reason` oneof — DisputeTransaction filing refused").optional(), "domain": z.string().describe("Stable grouping for the failing surface, e.g. \"fora.v1.ExchangeService\".\n Mirrors google.rpc.ErrorInfo.domain so generic tooling can group errors.").default(""), "domain_verification_failure": z.object({ "reason": z.enum(["DOMAIN_VERIFICATION_FAILURE_REASON_CHALLENGE_NOT_FOUND","DOMAIN_VERIFICATION_FAILURE_REASON_CHALLENGE_MISMATCH","DOMAIN_VERIFICATION_FAILURE_REASON_CHALLENGE_EXPIRED","DOMAIN_VERIFICATION_FAILURE_REASON_FETCH_FAILED","DOMAIN_VERIFICATION_FAILURE_REASON_EXCHANGE_NOT_AUTHORIZED","DOMAIN_VERIFICATION_FAILURE_REASON_KEY_REGISTRATION_FAILED","DOMAIN_VERIFICATION_FAILURE_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("The failure reason (defined-only, non-zero)") }).describe("`reason` oneof — domain verification failed").optional(), "message": z.string().describe("Developer-facing, NON-authoritative human message. Clients MUST branch on\n the typed reason below, never on this text. Servers SHOULD NOT place secrets,\n PII, or existence/authorization detail here that the closed typed reason\n deliberately withholds: unlike the enum, this free text is unbounded and\n easily becomes an existence oracle or leak channel (see `metadata`).").default(""), "metadata": z.record(z.string(), z.string()).describe("Dynamic key/value context that also appears in `message` (ids, limits,\n axes). Mirrors google.rpc.ErrorInfo.metadata. Strongly-typed context rides\n in the per-domain reason block below instead. Same leakage rule as `message`:\n servers SHOULD NOT put secrets, PII, or withheld existence/authorization\n detail here — it is the same potential side channel as the absence oracle.").optional(), "registration_failure": z.object({ "field_errors": z.array(z.object({ "error": z.string().min(1).max(255).describe("Developer-facing, NON-authoritative description of what failed\n (e.g. \"required\", \"must match ^[A-Z]{2}[0-9]+$\"). Wording is\n validator-defined and not stable across Exchanges; clients branch on\n `reason`, never on this text. States the constraint, NEVER the submitted\n value — the ErrorDetail leakage rule applies here too."), "path": z.string().max(255).describe("RFC 6901 JSON Pointer to the offending member, relative to\n registration_data (e.g. \"/vat_id\", \"/address/postal_code\"). The empty\n string addresses registration_data itself, for whole-object failures\n (oneOf, minProperties) that belong to no single member.").default("") }).describe("RegistrationFieldError — one registration_data member that failed the\n Exchange's published schema (AccountRegistration.data_schema).")).max(64).describe("When reason = INVALID_REGISTRATION_DATA: the registration_data members\n that are missing or do not conform. Empty for every other reason — enforced\n by the message rule above, not left to prose.").optional(), "reason": z.enum(["REGISTRATION_FAILURE_REASON_DOMAIN_NOT_VERIFIED","REGISTRATION_FAILURE_REASON_INVALID_KEY","REGISTRATION_FAILURE_REASON_SIGNATURE_INVALID","REGISTRATION_FAILURE_REASON_ALREADY_REGISTERED","REGISTRATION_FAILURE_REASON_QUOTA_EXCEEDED","REGISTRATION_FAILURE_REASON_INVALID_REGISTRATION_DATA","REGISTRATION_FAILURE_REASON_TERMS_DIGEST_STALE","REGISTRATION_FAILURE_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("The failure reason (defined-only, non-zero)") }).describe("`reason` oneof — agent/provider registration refused").optional(), "request_auth_failure": z.object({ "reason": z.enum(["REQUEST_AUTH_FAILURE_REASON_SIGNATURE_MISSING","REQUEST_AUTH_FAILURE_REASON_SIGNATURE_INVALID","REQUEST_AUTH_FAILURE_REASON_SIGNATURE_STALE"]).describe("The failure reason (defined-only, non-zero)") }).describe("`reason` oneof — RFC 9421 request signature on an RPC refused").optional(), "retrieval_auth_failure": z.object({ "reason": z.enum(["RETRIEVAL_AUTH_FAILURE_REASON_URL_EXPIRED","RETRIEVAL_AUTH_FAILURE_REASON_URL_SIGNATURE_MISSING","RETRIEVAL_AUTH_FAILURE_REASON_URL_EXPIRY_MISSING","RETRIEVAL_AUTH_FAILURE_REASON_URL_SIGNATURE_MISMATCH","RETRIEVAL_AUTH_FAILURE_REASON_AGENT_KEY_MISSING","RETRIEVAL_AUTH_FAILURE_REASON_PROOF_SIGNATURE_MISSING","RETRIEVAL_AUTH_FAILURE_REASON_KEYID_MISMATCH","RETRIEVAL_AUTH_FAILURE_REASON_THUMBPRINT_MISMATCH","RETRIEVAL_AUTH_FAILURE_REASON_PROOF_CREATED_MISSING","RETRIEVAL_AUTH_FAILURE_REASON_PROOF_EXPIRY_MISSING","RETRIEVAL_AUTH_FAILURE_REASON_PROOF_EXPIRED","RETRIEVAL_AUTH_FAILURE_REASON_PROOF_SIGNATURE_INVALID"]).describe("The failure reason (defined-only, non-zero)") }).describe("`reason` oneof — signed-URL / proof-of-possession check failed").optional(), "transaction_denial": z.object({ "exchange": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("Bare host of the Exchange that PRODUCED this denial, in the form \"Request\n recipient\" defines in the file header. Not an echo of what the caller sent:\n on a relayed or fanned-out execute the request went to a Broker, so the\n Exchange that refused may not be one the agent named (through\n BrokerService.ExecuteTransaction this detail reaches the agent inside\n UpstreamRefusal.detail, beside UpstreamRefusal.party). Carrying it here is\n what lets ACCOUNT_NOT_REGISTERED be actionable — the agent learns where to\n call Register without fetching a manifest to work it out. NOTHING SIGNS THIS\n VALUE: it rides in a response, and on a relayed path the response passed\n through an intermediary, so this field is exactly the unsigned addressing\n the request-side `exchange` field exists to refuse. Treat it as a HINT, not\n an instruction. Before acting on it — and registering is a consequential act,\n handing an operator's business data and a signed acceptance of that\n Exchange's terms to whoever answers — a caller MUST check the value against\n a domain it already trusts for this transaction: the signed `offer.exchange`\n of the items in the refused request, or its own RequestConstraints.exchanges\n set. A value matching neither is reported to the caller and never dialled,\n because a hostile intermediary that could choose it would be choosing where\n an unattended agent registers.").optional(), "offer_id": z.string().describe("DEPRECATED, never set. A whole-request refusal pertains to no one offer; a\n denial of one offer rides on that offer's TransactionResultItem. The field\n is retained because removing it would break the v1 wire contract; a\n receiver ignores it.").optional(), "reason": z.enum(["DENIAL_REASON_ACCOUNT_INACTIVE","DENIAL_REASON_INSUFFICIENT_BALANCE","DENIAL_REASON_RATE_LIMITED","DENIAL_REASON_CONTENT_UNAVAILABLE","DENIAL_REASON_RESTRICTION_NOT_SATISFIED","DENIAL_REASON_REPORTING_OVERDUE","DENIAL_REASON_OFFER_EXPIRED","DENIAL_REASON_SIGNATURE_INVALID","DENIAL_REASON_QUOTA_EXCEEDED","DENIAL_REASON_DELEGATION_INVALID","DENIAL_REASON_SCOPE_INSUFFICIENT","DENIAL_REASON_ENTITLEMENT_MISSING","DENIAL_REASON_ENTITLEMENT_MALFORMED","DENIAL_REASON_ENTITLEMENT_EXPIRED","DENIAL_REASON_ENTITLEMENT_WRONG_BUYER","DENIAL_REASON_SUBSCRIPTION_LAPSED","DENIAL_REASON_ENTITLEMENT_NOT_GRANTED","DENIAL_REASON_ACCOUNT_NOT_REGISTERED","DENIAL_REASON_RELAY_NOT_ACCEPTED","DENIAL_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("The denial reason (defined-only, non-zero)"), "restriction_mismatches": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("DEPRECATED, never set. RESTRICTION_NOT_SATISFIED is never sent: the\n Exchange never enforces a restriction (see Restriction). The field is\n retained because removing it would break the v1 wire contract; a receiver\n ignores it.").optional() }).describe("`reason` oneof — ExecuteTransaction refused as a whole request").optional(), "usage_report_rejection": z.object({ "reason": z.enum(["USAGE_REPORT_REJECTION_REASON_TRANSACTION_NOT_FOUND","USAGE_REPORT_REJECTION_REASON_DUPLICATE","USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED","USAGE_REPORT_REJECTION_REASON_MISSING_REQUIRED_FIELDS","USAGE_REPORT_REJECTION_REASON_MALFORMED","USAGE_REPORT_REJECTION_REASON_UNKNOWN_CRITICAL_EXTENSION"]).describe("The rejection reason (defined-only, non-zero)") }).describe("`reason` oneof — ReportUsage filing rejected").optional() }).describe("The party's typed reason, unchanged — the ErrorDetail it attached to its\n refusal. Absent when the party attached none or gave no answer.").optional(), "party": z.string().regex(new RegExp("^[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?(\\.[A-Za-z0-9]([A-Za-z0-9-]*[A-Za-z0-9])?)*(:(6553[0-5]|655[0-2][0-9]|65[0-4][0-9]{2}|6[0-4][0-9]{3}|[1-5][0-9]{4}|[1-9][0-9]{0,3}))?$")).max(260).describe("Bare host of the party that refused the call made on the caller's behalf,\n in the form \"Request recipient\" defines in the file header.") }).describe("Set only on a BrokerTransactionResponse, and only when the Exchange that\n owns this item refused the Broker's whole sub-request instead of answering\n it: a non-OK answer, or no answer at all. offer_id names the item;\n transaction_id, billing_id, cost, denial_reason, retrieval_endpoint and the\n other result fields stay unset. Every item the Broker sent in that\n sub-request carries the same refusal. Whether the item was purchased\n depends on the refusal's code. With the code \"unavailable\" or\n \"deadline_exceeded\" the outcome is unknown: the Exchange may have completed\n the purchase before the answer was lost, whether it gave no answer or\n answered that code, and retrying the agent's request with the same\n idempotency_key settles it, on any route: through the same Broker, another\n one, or directly (see UpstreamRefusal.code). With any other code the Exchange answered with a\n refusal that decided no item, so the item was not purchased. An Exchange\n never sets this field, and a Broker never replaces it with a denial_reason:\n an upstream refusal is not a per-item access decision.").optional(), "reporting_obligation": z.object({ "endpoint": z.string().describe("URL to submit the usage report to (if different from Exchange).").optional(), "ext": z.record(z.string(), z.any()).describe("Extension point").optional(), "ext_critical": z.array(z.string()).describe("Critical extension keys (COSE crit pattern, RFC 9052).\n Lists keys within ext that the consumer MUST understand.\n Unknown keys in this list → reject with UNKNOWN_CRITICAL_EXTENSION.\n Empty (default) → all ext keys are safe to ignore.").optional(), "required": z.boolean().describe("Whether post-usage reporting is required.").default(false), "required_fields": z.array(z.string()).describe("Field names that must be present in the report.").optional(), "window": z.string().describe("Duration within which the report must be submitted (e.g. \"86400s\" = 24\n hours; proto-JSON encodes Duration as seconds). A report that arrives after\n the window closed is refused with USAGE_REPORT_REJECTION_REASON_WINDOW_EXPIRED\n (Connect code failed_precondition).").optional() }).describe("Reporting requirements for this item.").optional(), "resource_title": z.string().describe("Resource title echoed from the Offer.").optional(), "restriction_mismatches": z.array(z.enum(["RESTRICTION_KIND_FUNCTION","RESTRICTION_KIND_GEOGRAPHY","RESTRICTION_KIND_USER_TYPE","RESTRICTION_KIND_OTHER"])).describe("DEPRECATED, never set. It named the restriction axes a request failed under\n DENIAL_REASON_RESTRICTION_NOT_SATISFIED, which is never sent: the Exchange\n never enforces a restriction (see Restriction). The field is retained\n because removing it would break the v1 wire contract; a receiver ignores it.").optional(), "retrieval_endpoint": z.string().describe("Signed retrieval URL for this item. Bound to the requesting agent's identity\n via the parent TransactionResponse.agent_identity_hash (shared across all\n batch items) — on a BrokerTransactionResponse, the agent_identity_hash of\n the ExchangeOutcome for this item's Exchange; expires at expires_at. Absent if this item was denied or its\n delivery_method is not signed-URL-based.").optional(), "subscription_id": z.string().describe("If under subscription, no per-request charge.").optional(), "subscription_unit_value": z.object({ "amount": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Exact decimal string (not a float), e.g. \"19.99\". Denominated in `currency`.").default(""), "currency": z.string().default(""), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).optional() }).describe("Computed per-unit cost for financial attribution on subscription transactions.\n Even when cost.amount=\"0\" (subscription), this field carries the value\n of the access for accounting purposes (e.g., ASC 606 prepaid drawdown).").optional(), "transaction_id": z.string().describe("Exchange-assigned transaction identifier.").default("") }).describe("TransactionResultItem — Result for a single offer in a batch transaction.")).describe("Per-offer results (one entry per committed item, in original order).").optional(), "subscription_quota": z.array(z.object({ "quota_limit": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Total allowed in the current period.").optional(), "quota_remaining": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Remaining in the current period.").optional(), "quota_used": z.coerce.number().int().gte(-2147483648).lte(2147483647).describe("Used so far in the current period.").optional(), "resets_at": z.string().datetime({ offset: true }).describe("When the quota counter resets (UTC).").optional(), "subscription_id": z.string().describe("Subscription this quota applies to.").default(""), "unit": z.string().describe("What is being metered. Distinguishes access count quotas from\n spend quotas from burst limits.\n Standard values: \"accesses\", \"tokens\", \"spend_cents\", \"burst\"").optional() }).describe("SubscriptionQuotaInfo — Proactive quota signaling for subscription access.\n\nAnalogous to RateLimitInfo (which signals API request rate limits), this\n signals subscription consumption quotas. Enables agents to throttle\n proactively instead of discovering exhaustion via denial.\n\n Returned on Offer (per-offer quota visibility) and TransactionResponse\n (post-transaction remaining quota). A subscription may have multiple\n independent quotas (access count + spend cap + burst limit), so this\n message is used as a repeated field.\n\n Quota decrement timing: the counter increments at ExecuteTransaction\n (optimistic decrement, before delivery). If delivery fails, the agent\n files a DisputeTransaction which may reverse the decrement. This is\n consistent with the billing model (billing_id created at transaction time).")).describe("Post-transaction quota state. Tells the agent how much quota remains\n after this transaction. Enables proactive throttling (\"1 access left\").\n Multiple entries for multi-dimensional quotas.").optional(), "total_cost": z.object({ "amount": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).describe("Exact decimal string (not a float), e.g. \"19.99\". Denominated in `currency`.").default(""), "currency": z.string().default(""), "unit_cost": z.string().regex(new RegExp("^([0-9]+([.][0-9]+)?)?$")).max(32).optional() }).describe("Aggregate cost of the purchased items (those without denial_reason): their\n cost.amount summed as exact decimals, in their one shared currency.\n unit_cost is unset. UNSET when the purchased items span more than one\n currency: amounts are never summed across currencies, and currency\n conversion is out of scope for this version. A caller that needs the totals\n then sums each item's cost per currency itself.").optional(), "ver": z.string().describe("FORA protocol version — \"1.0\". Stamped by the sender from a single\n constant; advisory on receive. See \"Protocol version\" in the file header.").default("") }).describe("TransactionResponse — Exchange confirms the transaction(s).\n\nItems-only: every per-result datum lives in `items`\n (one TransactionResultItem per committed offer, in original order); the\n top-level fields carry only the shared aggregate state. A single offer is the\n degenerate 1-element `items`.\n\n Every per-item denial is in the body, on TransactionResultItem.denial_reason,\n whatever the item count. A one-item purchase that is denied is this response\n with its one item denied; it is never turned into a non-OK error. Only a\n refusal of the whole request, which decides no item, is a non-OK error\n carrying ErrorDetail.transaction_denial (see TransactionDenial).")))();
diff --git a/proto/CHANGELOG.md b/proto/CHANGELOG.md
index 6d248c8c..745ccc84 100644
--- a/proto/CHANGELOG.md
+++ b/proto/CHANGELOG.md
@@ -2,6 +2,28 @@
## Unreleased
+**A Broker orders offers by relevance, and the agent decides (docs and comments only; no wire
+change).** The docs no longer describe the Broker as choosing the cheapest offer or the best deal.
+A Broker returns every Exchange's signed offers in its order of relevance to the agent's request,
+optionally narrowed to the kinds of license the agent accepts (`acceptable_restrictions`). It
+groups offers for the same content without dropping any, and it never buys, resells or puts one
+offer in place of another. Price is not a signal: each offer carries its price and
+`Pricing.unit_cost`, and the agent compares them under its own policy. The Selection Engine page
+is rewritten on these lines, and the Content Deduplication, Discovery Paths, walkthrough, Agent
+SDK, money-flow and operator pages follow. Two comments in `fora.proto` (`Offer.pricing` and the
+one-term-per-offer note) now say an agent compares the price, not that a Broker ranks on it.
+
+**The edge function refuses AI search crawlers by default (docs only; no wire change).** Bot
+Detection now describes the lists FORA Server ships: AI search crawlers such as PerplexityBot
+get a `403` on licensed paths by default, and a provider changes the built-in lists with
+`BOT_UA_ALLOW_JSON` and `BOT_UA_DENY_JSON`. The `bot_detection.ai_search_policy` setting the page
+described doesn't exist and is gone. Googlebot, Bingbot and Applebot are never refused. The
+publishers page, the AI crawlers resource, the bot table and the rules generator say the same.
+
+**Publisher onboarding names the console's operator (docs only).** FORA is open-source software,
+not an operator. The publisher console on fora-protocol.org is run by Postindustria, which
+charges publishers no Exchange fee beyond the payment system's transaction fees.
+
**Proto comments state the SSRF guards instead of citing the threat model (comments only; no
wire change).** The `License.uri` comment and the CoMP comments on `Package.licenseurl`,
`Package.reporturl` and `Retrieval.endpoint` cited threat ID T-LIC-1 in the threat model. They
diff --git a/proto/fora/v1/fora.proto b/proto/fora/v1/fora.proto
index 1b054687..d182af38 100644
--- a/proto/fora/v1/fora.proto
+++ b/proto/fora/v1/fora.proto
@@ -930,8 +930,8 @@ message Offer {
// licensing arrangement: each projected LicenseTerm yields its own offer, and
// the Exchange moves that term's price here, so the term inside `terms`
// carries no pricing of its own (the offer.terms.pricing_unset rule). This is
- // the price execute charges, the price a Broker ranks and compares across
- // Exchanges, and the price a metered purchase is charged at. A resource with
+ // the price execute charges, the price an agent compares across Exchanges,
+ // and the price a metered purchase is charged at. A resource with
// multiple alternative terms (e.g. dual-licensed) produces multiple separate
// offers, one per term — never one offer with a "headline" picked among them.
//
@@ -1148,9 +1148,9 @@ message Offer {
// rule instead. `ResourceEntry.terms` is the plural side, bounded at 32: a
// resource carries many terms, an offer sells one of them.
//
- // The distinction is not cosmetic. `pricing` is what a Broker ranks on and
- // what execute charges, so a term that is not on its own offer has no price
- // a Broker can compare and no offer_id an agent can buy. Fusing several
+ // The distinction is not cosmetic. `pricing` is what an agent sees and what
+ // execute charges, so a term that is not on its own offer has no price an
+ // agent can see and no offer_id an agent can buy. Fusing several
// terms onto one offer makes every term but the first unsellable — a
// dual-licensed resource sells only under whichever arrangement the
// publisher happened to store first.
diff --git a/scripts/check-doc-conformance.sh b/scripts/check-doc-conformance.sh
index b7646a67..62e301e5 100755
--- a/scripts/check-doc-conformance.sh
+++ b/scripts/check-doc-conformance.sh
@@ -317,11 +317,16 @@ fi
# --- 3. Wrong-org references -----------------------------------------------
# The project lives at github.com/FORA-Protocol. The pre-rename org literal must not
-# appear anywhere — code, docs, config, copyright. The literal is built from two
-# pieces below so this script does not contain it verbatim, and therefore scans its
-# own file too (no self-exclusion blind spot).
+# appear anywhere — code, docs, config, copyright. The match is case-sensitive: an org,
+# a module path or a package name is written in lower case, while the company's name in
+# prose ("Postindustria runs the console on this site") is a fact about an operator,
+# not a stale reference. The literal is built from two pieces below so this script
+# does not contain it verbatim, and therefore scans its own file too (no
+# self-exclusion blind spot).
old_org='postin''dustria'
-org_hits=$(git grep -niF "$old_org" -- . 2>/dev/null || true)
+# The company's own website, postindustria.com, is a link to the company, not a
+# stale org reference, so a match followed by ".com" is allowed.
+org_hits=$(git grep -nE "${old_org}([^.]|\.[^c]|\.c[^o]|\.co[^m]|$)" -- . 2>/dev/null || true)
if [ -n "$org_hits" ]; then
echo "::error::stale '${old_org}' org reference (the project is github.com/FORA-Protocol):"
echo "$org_hits"
diff --git a/website/brand/fora-brand/SKILL.md b/website/brand/fora-brand/SKILL.md
index 03643190..f6312ffa 100644
--- a/website/brand/fora-brand/SKILL.md
+++ b/website/brand/fora-brand/SKILL.md
@@ -116,6 +116,7 @@ Read `reference/messaging.md` before writing any copy. The short version:
- Write **FORA** in prose. "The FORA protocol" or "FORA Protocol" when you need
the noun. The lowercase "fora protocol" exists only inside the logo.
- It stands for **Federated Open Resource Access**. Spell it out once, if at all.
+ *Fora* is also the plural of *forum*: the mark lays the plan of a Roman forum over a wax seal.
- Never "Fora" or "F.O.R.A.".
- FORA is an **open protocol**, not an "industry standard" and not adopted by any
standards body.
diff --git a/website/brand/fora-brand/reference/messaging.md b/website/brand/fora-brand/reference/messaging.md
index ee6951bf..471d8d47 100644
--- a/website/brand/fora-brand/reference/messaging.md
+++ b/website/brand/fora-brand/reference/messaging.md
@@ -80,7 +80,8 @@ Dispute), as the home page and /music/ do.
## Facts you may state
-- FORA stands for Federated Open Resource Access.
+- FORA stands for Federated Open Resource Access, and *fora* is also the plural of *forum*:
+ the mark lays the plan of a Roman forum over a wax seal.
- The protocol specification and the reference implementation are open source
under Apache 2.0. The documentation is CC BY-ND 4.0.
- SDKs exist for Go, TypeScript and Python. Agents can also use FORA through an
diff --git a/website/public/brand/SKILL.md b/website/public/brand/SKILL.md
index 03643190..f6312ffa 100644
--- a/website/public/brand/SKILL.md
+++ b/website/public/brand/SKILL.md
@@ -116,6 +116,7 @@ Read `reference/messaging.md` before writing any copy. The short version:
- Write **FORA** in prose. "The FORA protocol" or "FORA Protocol" when you need
the noun. The lowercase "fora protocol" exists only inside the logo.
- It stands for **Federated Open Resource Access**. Spell it out once, if at all.
+ *Fora* is also the plural of *forum*: the mark lays the plan of a Roman forum over a wax seal.
- Never "Fora" or "F.O.R.A.".
- FORA is an **open protocol**, not an "industry standard" and not adopted by any
standards body.
diff --git a/website/public/brand/fora-brand.zip b/website/public/brand/fora-brand.zip
index 160ca9ee..755f75b0 100644
Binary files a/website/public/brand/fora-brand.zip and b/website/public/brand/fora-brand.zip differ
diff --git a/website/public/brand/messaging.md b/website/public/brand/messaging.md
index ee6951bf..471d8d47 100644
--- a/website/public/brand/messaging.md
+++ b/website/public/brand/messaging.md
@@ -80,7 +80,8 @@ Dispute), as the home page and /music/ do.
## Facts you may state
-- FORA stands for Federated Open Resource Access.
+- FORA stands for Federated Open Resource Access, and *fora* is also the plural of *forum*:
+ the mark lays the plan of a Roman forum over a wax seal.
- The protocol specification and the reference implementation are open source
under Apache 2.0. The documentation is CC BY-ND 4.0.
- SDKs exist for Go, TypeScript and Python. Agents can also use FORA through an
diff --git a/website/public/resources/agent-identity/id-1.png b/website/public/resources/agent-identity/id-1.png
new file mode 100644
index 00000000..f477388c
Binary files /dev/null and b/website/public/resources/agent-identity/id-1.png differ
diff --git a/website/public/resources/content-marketplaces/cloudflare/cf-1.png b/website/public/resources/content-marketplaces/cloudflare/cf-1.png
new file mode 100644
index 00000000..3b51a544
Binary files /dev/null and b/website/public/resources/content-marketplaces/cloudflare/cf-1.png differ
diff --git a/website/public/resources/content-marketplaces/cloudflare/cf-2.png b/website/public/resources/content-marketplaces/cloudflare/cf-2.png
new file mode 100644
index 00000000..eedbd93a
Binary files /dev/null and b/website/public/resources/content-marketplaces/cloudflare/cf-2.png differ
diff --git a/website/public/resources/content-marketplaces/hub/hub-2.png b/website/public/resources/content-marketplaces/hub/hub-2.png
new file mode 100644
index 00000000..ddcdf29e
Binary files /dev/null and b/website/public/resources/content-marketplaces/hub/hub-2.png differ
diff --git a/website/public/resources/content-marketplaces/hub/hub-3.png b/website/public/resources/content-marketplaces/hub/hub-3.png
new file mode 100644
index 00000000..b96382d5
Binary files /dev/null and b/website/public/resources/content-marketplaces/hub/hub-3.png differ
diff --git a/website/public/resources/content-marketplaces/hub/hub-5.png b/website/public/resources/content-marketplaces/hub/hub-5.png
new file mode 100644
index 00000000..75e2fbda
Binary files /dev/null and b/website/public/resources/content-marketplaces/hub/hub-5.png differ
diff --git a/website/public/resources/content-marketplaces/microsoft-pcm/pcm-1.png b/website/public/resources/content-marketplaces/microsoft-pcm/pcm-1.png
new file mode 100644
index 00000000..b5cbec38
Binary files /dev/null and b/website/public/resources/content-marketplaces/microsoft-pcm/pcm-1.png differ
diff --git a/website/public/resources/content-marketplaces/microsoft-pcm/pcm-3.png b/website/public/resources/content-marketplaces/microsoft-pcm/pcm-3.png
new file mode 100644
index 00000000..06d931fc
Binary files /dev/null and b/website/public/resources/content-marketplaces/microsoft-pcm/pcm-3.png differ
diff --git a/website/public/resources/content-marketplaces/perplexity/perplexity-1.png b/website/public/resources/content-marketplaces/perplexity/perplexity-1.png
new file mode 100644
index 00000000..9da28622
Binary files /dev/null and b/website/public/resources/content-marketplaces/perplexity/perplexity-1.png differ
diff --git a/website/public/resources/content-marketplaces/perplexity/perplexity-2.png b/website/public/resources/content-marketplaces/perplexity/perplexity-2.png
new file mode 100644
index 00000000..8b07a7e0
Binary files /dev/null and b/website/public/resources/content-marketplaces/perplexity/perplexity-2.png differ
diff --git a/website/public/resources/content-marketplaces/perplexity/perplexity-3.png b/website/public/resources/content-marketplaces/perplexity/perplexity-3.png
new file mode 100644
index 00000000..2d02baeb
Binary files /dev/null and b/website/public/resources/content-marketplaces/perplexity/perplexity-3.png differ
diff --git a/website/public/resources/content-marketplaces/prorata-gist/prorata-1.png b/website/public/resources/content-marketplaces/prorata-gist/prorata-1.png
new file mode 100644
index 00000000..6fb82798
Binary files /dev/null and b/website/public/resources/content-marketplaces/prorata-gist/prorata-1.png differ
diff --git a/website/public/resources/content-marketplaces/prorata-gist/prorata-2.png b/website/public/resources/content-marketplaces/prorata-gist/prorata-2.png
new file mode 100644
index 00000000..68abccca
Binary files /dev/null and b/website/public/resources/content-marketplaces/prorata-gist/prorata-2.png differ
diff --git a/website/public/resources/content-marketplaces/prorata-gist/prorata-4.png b/website/public/resources/content-marketplaces/prorata-gist/prorata-4.png
new file mode 100644
index 00000000..0231d3d0
Binary files /dev/null and b/website/public/resources/content-marketplaces/prorata-gist/prorata-4.png differ
diff --git a/website/public/resources/content-marketplaces/tollbit/tollbit-1.png b/website/public/resources/content-marketplaces/tollbit/tollbit-1.png
new file mode 100644
index 00000000..8e2f5c6b
Binary files /dev/null and b/website/public/resources/content-marketplaces/tollbit/tollbit-1.png differ
diff --git a/website/public/resources/content-marketplaces/tollbit/tollbit-3.png b/website/public/resources/content-marketplaces/tollbit/tollbit-3.png
new file mode 100644
index 00000000..a589fc7b
Binary files /dev/null and b/website/public/resources/content-marketplaces/tollbit/tollbit-3.png differ
diff --git a/website/public/resources/content-marketplaces/tollbit/tollbit-4.png b/website/public/resources/content-marketplaces/tollbit/tollbit-4.png
new file mode 100644
index 00000000..08f2432a
Binary files /dev/null and b/website/public/resources/content-marketplaces/tollbit/tollbit-4.png differ
diff --git a/website/public/resources/standards-map/standards-1.png b/website/public/resources/standards-map/standards-1.png
new file mode 100644
index 00000000..4a3a65cb
Binary files /dev/null and b/website/public/resources/standards-map/standards-1.png differ
diff --git a/website/public/resources/standards-map/standards-3.png b/website/public/resources/standards-map/standards-3.png
new file mode 100644
index 00000000..2ee0a6a7
Binary files /dev/null and b/website/public/resources/standards-map/standards-3.png differ
diff --git a/website/public/resources/standards-map/standards-5.png b/website/public/resources/standards-map/standards-5.png
new file mode 100644
index 00000000..20e080e1
Binary files /dev/null and b/website/public/resources/standards-map/standards-5.png differ
diff --git a/website/public/resources/web-bot-auth/wba-1.png b/website/public/resources/web-bot-auth/wba-1.png
new file mode 100644
index 00000000..f208aaac
Binary files /dev/null and b/website/public/resources/web-bot-auth/wba-1.png differ
diff --git a/website/public/resources/web-bot-auth/wba-10.png b/website/public/resources/web-bot-auth/wba-10.png
new file mode 100644
index 00000000..ab89ba8f
Binary files /dev/null and b/website/public/resources/web-bot-auth/wba-10.png differ
diff --git a/website/public/resources/web-bot-auth/wba-4.png b/website/public/resources/web-bot-auth/wba-4.png
new file mode 100644
index 00000000..5195a9ea
Binary files /dev/null and b/website/public/resources/web-bot-auth/wba-4.png differ
diff --git a/website/public/resources/web-bot-auth/wba-5.png b/website/public/resources/web-bot-auth/wba-5.png
new file mode 100644
index 00000000..52a98bba
Binary files /dev/null and b/website/public/resources/web-bot-auth/wba-5.png differ
diff --git a/website/public/resources/web-bot-auth/wba-6.png b/website/public/resources/web-bot-auth/wba-6.png
new file mode 100644
index 00000000..a4c014c0
Binary files /dev/null and b/website/public/resources/web-bot-auth/wba-6.png differ
diff --git a/website/public/resources/web-bot-auth/wba-8.png b/website/public/resources/web-bot-auth/wba-8.png
new file mode 100644
index 00000000..d546a2c8
Binary files /dev/null and b/website/public/resources/web-bot-auth/wba-8.png differ
diff --git a/website/public/resources/web-bot-auth/wba-9.png b/website/public/resources/web-bot-auth/wba-9.png
new file mode 100644
index 00000000..588ba57b
Binary files /dev/null and b/website/public/resources/web-bot-auth/wba-9.png differ
diff --git a/website/redirects.json b/website/redirects.json
index 3789720f..98681b86 100644
--- a/website/redirects.json
+++ b/website/redirects.json
@@ -4,6 +4,31 @@
"status": "302",
"target": "https://fora-protocol.org"
},
+ {
+ "source": "/getting-started/industry-landscape/",
+ "target": "/resources/standards-map/",
+ "status": "301"
+ },
+ {
+ "source": "/components/broker/selection-engine/",
+ "target": "/components/broker/overview/",
+ "status": "301"
+ },
+ {
+ "source": "/components/broker/content-dedup/",
+ "target": "/components/broker/overview/",
+ "status": "301"
+ },
+ {
+ "source": "/components/broker/budget-enforcement/",
+ "target": "/components/broker/overview/",
+ "status": "301"
+ },
+ {
+ "source": "/components/broker/deployment/",
+ "target": "/components/broker/overview/",
+ "status": "301"
+ },
{
"source": "/security/threat-model/",
"target": "/protocol/authentication/",
diff --git a/website/routes.txt b/website/routes.txt
index fdae2f45..76e54cd1 100644
--- a/website/routes.txt
+++ b/website/routes.txt
@@ -5,6 +5,7 @@
# this repository links to it, that link breaks. Retiring one therefore needs
# a rule in redirects.json — see docs/retiring-a-url.md.
/
+/about/
/architecture/deployment-models/
/architecture/nfr/
/architecture/production-architecture/
@@ -12,11 +13,7 @@
/components/agent-sdk/budget-reporting/
/components/agent-sdk/fetch-flow/
/components/agent-sdk/overview/
-/components/broker/budget-enforcement/
-/components/broker/content-dedup/
-/components/broker/deployment/
/components/broker/overview/
-/components/broker/selection-engine/
/components/content-ingestion/catalog-compilation/
/components/content-ingestion/normalization/
/components/content-ingestion/overview/
@@ -50,7 +47,6 @@
/getting-started/for-providers/
/getting-started/for-verification-vendors/
/getting-started/how-money-flows/
-/getting-started/industry-landscape/
/getting-started/live-demo-music/
/getting-started/poc-walkthrough/
/getting-started/publisher-onboarding/
@@ -92,7 +88,16 @@
/reference/proto-fora/
/reference/standards/
/resources/
+/resources/agent-identity/
/resources/ai-crawlers/
+/resources/content-marketplaces/
+/resources/content-marketplaces/cloudflare/
+/resources/content-marketplaces/microsoft-pcm/
+/resources/content-marketplaces/perplexity/
+/resources/content-marketplaces/prorata-gist/
+/resources/content-marketplaces/tollbit/
/resources/music-dsp/
+/resources/standards-map/
/resources/transparent-collective/
/resources/transparent-collective/evidence/
+/resources/web-bot-auth/
diff --git a/website/src/components/SiteNav.astro b/website/src/components/SiteNav.astro
index 6b3d108d..d13120a5 100644
--- a/website/src/components/SiteNav.astro
+++ b/website/src/components/SiteNav.astro
@@ -108,6 +108,13 @@ const { current } = Astro.props;
}
.nav-links a:hover { color: var(--off-white); }
.nav-links a.on { color: var(--off-white); }
+ .nav-links a { white-space: nowrap; }
+
+ /* Just above the menu breakpoint, the full row of links fits only with
+ tighter gaps. */
+ @media (min-width: 769px) and (max-width: 960px) {
+ .nav-links { gap: 1.25rem; }
+ }
.menu-btn, .nav-spacer { display: none; }
diff --git a/website/src/components/landing/AudienceHero.astro b/website/src/components/landing/AudienceHero.astro
index c16ae690..a4e464d5 100644
--- a/website/src/components/landing/AudienceHero.astro
+++ b/website/src/components/landing/AudienceHero.astro
@@ -7,6 +7,7 @@ import type { LandingPageKey } from '../../consts';
interface Props {
current: LandingPageKey;
+ /** May hold a link, such as a resource page's parent topic. */
badge: string;
/** The heading; may hold and one phrase, shown in blue. */
title: string;
@@ -22,7 +23,7 @@ const hasAside = Astro.slots.has('aside');
-
{badge}
+
{sub}
{updated &&
Updated
}
@@ -36,6 +37,8 @@ const hasAside = Astro.slots.has('aside');
.audience-hero { min-height: 0; }
.audience-hero .hero-content { padding: 5rem 2rem; }
.audience-text { display: flex; flex-direction: column; align-items: center; }
+ .hero-badge :global(a) { color: inherit; text-decoration: underline; text-decoration-thickness: 1px; text-underline-offset: 3px; text-decoration-color: #8192ca; }
+ .hero-badge :global(a:hover) { text-decoration-color: currentColor; }
.hero-updated { margin: -0.75rem 0 1.75rem; font-family: var(--mono); font-size: 0.8125rem; letter-spacing: 0.04em; color: var(--gray, #8a8aa0); }
.hero-content.with-aside {
diff --git a/website/src/components/landing/Band.astro b/website/src/components/landing/Band.astro
index f1a36aab..32b51c19 100644
--- a/website/src/components/landing/Band.astro
+++ b/website/src/components/landing/Band.astro
@@ -4,8 +4,14 @@
// in a column beside it (an offer card, a code card).
//
// The tone sets the colour variables that the components inside read
-// (--lp-heading, --lp-text, --lp-label, --lp-strong, --lp-link, --lp-rule), so
-// a component looks right on a dark or a light band without knowing which.
+// (--lp-heading, --lp-text, --lp-label, --lp-strong, --lp-link, --lp-rule,
+// --lp-muted, --lp-fora, --lp-bg, --lp-surface), so a component looks right on
+// a dark or a light band without knowing which. --lp-bg is the band itself;
+// --lp-surface is a card on it; --lp-muted is secondary text; --lp-fora is
+// FORA's blue, the one colour that means FORA.
+//
+// `compact` halves the vertical padding, for a strip (KeyFacts) that breaks
+// the page up rather than opening a section.
interface Props {
tone?: 'black' | 'near' | 'light';
label?: string;
@@ -15,21 +21,28 @@ interface Props {
id?: string;
/** Width of the aside column, when there is one. */
asideWidth?: string;
+ /** How the aside lines up with the text column: centred (the default) or at the top. */
+ asideAlign?: 'center' | 'start';
+ compact?: boolean;
}
-const { tone = 'black', label, title, lead, id, asideWidth = '1fr' } = Astro.props;
+const { tone = 'black', label, title, lead, id, asideWidth = '1fr', asideAlign = 'center', compact = false } = Astro.props;
const hasAside = Astro.slots.has('aside');
+// A band with no header (only components, such as the About FORA card) renders no empty header block.
+const hasHeader = Boolean(label || title || lead || hasAside || Astro.slots.has('intro'));
---
-
+
-
-
- {label &&
{label}
}
- {title &&
{title}
}
- {lead && }
-
+ {hasHeader && (
+
+
+ {label &&
{label}
}
+ {title &&
{title}
}
+ {lead && }
+
+
+ {hasAside &&
}
- {hasAside &&
}
-
+ )}
@@ -43,9 +56,11 @@ const hasAside = Astro.slots.has('aside');
--lp-strong: var(--white);
--lp-link: var(--accent-text);
--lp-rule: var(--mid);
+ --lp-muted: var(--gray);
+ --lp-fora: var(--accent-text);
}
- .band-black { background: var(--black); border-top: 1px solid var(--mid); }
- .band-near { background: var(--near-black); }
+ .band-black { --lp-bg: var(--black); --lp-surface: var(--near-black); background: var(--black); border-top: 1px solid var(--mid); }
+ .band-near { --lp-bg: var(--near-black); --lp-surface: var(--black); background: var(--near-black); }
.band-light {
--lp-heading: var(--black);
--lp-text: #444456;
@@ -53,6 +68,10 @@ const hasAside = Astro.slots.has('aside');
--lp-strong: var(--black);
--lp-link: var(--accent-dark);
--lp-rule: #e4e4ec;
+ --lp-muted: #6b6b80;
+ --lp-fora: var(--accent-dark);
+ --lp-bg: var(--bg-light);
+ --lp-surface: #fff;
background: var(--bg-light);
color: var(--near-black);
}
@@ -63,7 +82,9 @@ const hasAside = Astro.slots.has('aside');
gap: 4rem;
align-items: center;
}
+ .band-pair-start { align-items: start; }
.band-pair:not(:last-child) { margin-bottom: 3.5rem; }
+ .band.band-compact { padding-top: 3.5rem; padding-bottom: 3.5rem; }
.band-label {
font-size: 0.875rem;
font-weight: 600;
@@ -82,6 +103,7 @@ const hasAside = Astro.slots.has('aside');
}
.band-lead { font-size: 1.125rem; line-height: 1.7; color: var(--lp-text); max-width: 44rem; }
.band-lead :global(strong) { color: var(--lp-strong); font-weight: 600; }
+ .band-lead :global(a) { color: var(--lp-link); }
.band :global(a:focus-visible) { outline: 2px solid var(--accent-text); outline-offset: 2px; }
@media (max-width: 900px) {
@@ -89,5 +111,6 @@ const hasAside = Astro.slots.has('aside');
}
@media (max-width: 560px) {
.band { padding: 4rem 1rem; }
+ .band.band-compact { padding-top: 2.75rem; padding-bottom: 2.75rem; }
}
diff --git a/website/src/components/landing/BotTable.astro b/website/src/components/landing/BotTable.astro
index 170c73c5..9a78100c 100644
--- a/website/src/components/landing/BotTable.astro
+++ b/website/src/components/landing/BotTable.astro
@@ -6,7 +6,7 @@ import { AI_BOTS, CHECKED, type BotKind } from '../../data/ai-bots';
const GROUPS: { kind: BotKind; title: string; rule: string }[] = [
{ kind: 'training', title: 'Training and datasets', rule: 'Refused in robots.txt across the site, and a 403 on the paths you sell.' },
{ kind: 'live-fetch', title: 'Live fetch', rule: 'A 403 on the paths you sell, where they can buy. Many don\'t follow robots.txt.' },
- { kind: 'ai-search', title: 'AI search', rule: 'Your choice. Refusing one takes you out of that vendor\'s AI search.' },
+ { kind: 'ai-search', title: 'AI search', rule: 'Your choice in robots.txt: refusing one takes you out of that vendor\'s AI search. On the paths you sell, FORA\'s edge function refuses them unless you let them through.' },
{ kind: 'search', title: 'Search crawlers that also feed AI', rule: 'Never refused: they bring your search traffic. Their AI uses are refused with the tokens and page settings below.' },
{ kind: 'token', title: 'robots.txt tokens', rule: 'No crawler sends these names. They exist only to refuse a use in robots.txt.' },
];
diff --git a/website/src/components/landing/Callout.astro b/website/src/components/landing/Callout.astro
new file mode 100644
index 00000000..0e7dd8f9
--- /dev/null
+++ b/website/src/components/landing/Callout.astro
@@ -0,0 +1,183 @@
+---
+// A boxed aside whose tag says who it's for. The left edge says whose voice
+// it is: grey for an aside to one of your teams (engineers, legal, search),
+// blue for FORA speaking about itself (define, about, takeaway), amber for a
+// vendor's own warning (caution). Amber (#b07d2a) is the one colour the
+// editorial patterns add; text never uses it, and it holds at least 3.5:1
+// against every band. Reads the band's colour variables.
+//
+// Each kind has a default tag; `define` and `takeaway` need their own. The
+// body is, in order: the default slot, `text` (an HTML paragraph), `points`
+// (bold lead-in and text, as a list, or as two columns with `columns`), and
+// `links` (a row joined by " · "). `collapse` puts the body behind a
+// disclosure with that label on narrow screens (≤640px); wider, it is open
+// and the disclosure is hidden. Shared texts (About FORA, the definitions,
+// the search callout) come from src/data/content-marketplaces.ts.
+//
+//
+//
+//
+type Kind = 'engineers' | 'legal' | 'search' | 'define' | 'about' | 'takeaway' | 'caution';
+interface Props {
+ kind: Kind;
+ tag?: string;
+ /** An HTML paragraph, after the slot. */
+ text?: string;
+ /** Bold lead-in and HTML text per point; a point may be text only. */
+ points?: { title?: string; text: string }[];
+ /** Lay `points` out in two columns (About FORA). */
+ columns?: boolean;
+ links?: { href: string; text: string }[];
+ /** On narrow screens, the summary of a disclosure that holds the body. */
+ collapse?: string;
+ id?: string;
+}
+const DEFAULT_TAG: Partial> = {
+ engineers: 'For your engineers',
+ legal: 'For legal',
+ search: 'Search traffic',
+ about: 'About FORA',
+ caution: 'What can go wrong',
+};
+const VOICE: Record = {
+ engineers: 'team', legal: 'team', search: 'team',
+ define: 'fora', about: 'fora', takeaway: 'fora',
+ caution: 'vendor',
+};
+// 16px line icons, drawn in the tag's colour.
+const ICON: Record = {
+ engineers: '',
+ legal: '',
+ search: '',
+ define: '',
+ about: '',
+ takeaway: '',
+ caution: '',
+};
+
+const { kind, tag = DEFAULT_TAG[kind], text, points, columns = false, links, collapse, id } = Astro.props;
+if (!tag) throw new Error(`Callout kind="${kind}" needs a tag`);
+const voice = VOICE[kind];
+const Wrap = collapse ? 'details' : 'div';
+---
+
+
+
diff --git a/website/src/components/landing/CardGrid.astro b/website/src/components/landing/CardGrid.astro
index ccea2561..60338a66 100644
--- a/website/src/components/landing/CardGrid.astro
+++ b/website/src/components/landing/CardGrid.astro
@@ -1,13 +1,22 @@
---
-// A row of three short cards: a bold title and a sentence or two.
+// A row of three short cards: a bold title and a sentence or two. A card that
+// describes a design carries its `status` (a StatusTag above the title).
+// Reads the band's colour variables, so the cards sit on any tone: on a black
+// band they are near-black with a dark rule, as they always were.
+import StatusTag from './StatusTag.astro';
+import type { StatusKey } from '../../data/content-marketplaces';
+
interface Props {
- items: { title: string; text: string }[];
+ items: { title: string; text: string; status?: StatusKey }[];
}
const { items } = Astro.props;
---
{items.map((it) => (
-
{it.title}
{it.text}
+
+ {it.status &&
}
+ {it.title}
+
))}
@@ -15,12 +24,13 @@ const { items } = Astro.props;
.card-grid { display: grid; grid-template-columns: repeat(3, 1fr); gap: 1rem; margin-top: 3rem; }
.card {
padding: 1.5rem;
- border: 1px solid var(--mid);
+ border: 1px solid var(--lp-rule, var(--mid));
border-radius: 12px;
- background: var(--near-black);
+ background: var(--lp-surface, var(--near-black));
}
- b { display: block; color: var(--white); font-size: 1.125rem; margin-bottom: 0.5rem; }
- p { color: var(--light-gray); line-height: 1.6; }
+ .card-status { margin-bottom: 0.875rem; }
+ b { display: block; color: var(--lp-heading, var(--white)); font-size: 1.125rem; margin-bottom: 0.5rem; }
+ p { color: var(--lp-text, var(--light-gray)); line-height: 1.6; }
@media (max-width: 900px) {
.card-grid { grid-template-columns: 1fr; }
}
diff --git a/website/src/components/landing/CompareTable.astro b/website/src/components/landing/CompareTable.astro
new file mode 100644
index 00000000..d39fd7dc
--- /dev/null
+++ b/website/src/components/landing/CompareTable.astro
@@ -0,0 +1,78 @@
+---
+// The comparison table of a vendor page: one row per question, a column for
+// the vendor and one for FORA (the only blue heading). A vendor page builds
+// its rows with compareRows() from src/data/content-marketplaces.ts, so every
+// page asks the same eight questions in the same order. Cells may hold inline
+// HTML: links, , citation markers.
+//
+// Below 760px each row becomes a card titled with its question, its two
+// answers labelled, so the page never scrolls sideways. For more than two
+// columns, use Scorecard. Reads the band's variables.
+//
+//
+interface Props {
+ vendor: string;
+ rows: { q: string; cells: readonly [string, string] }[];
+ /** The table's accessible name. */
+ label?: string;
+}
+const { vendor, rows, label } = Astro.props;
+---
+
+
+
+
+
The question
{vendor}
FORA
+
+
+ {rows.map((r) => (
+
+
{r.q}
+
+
+
+ ))}
+
+
+
+
+
diff --git a/website/src/components/landing/Figure.astro b/website/src/components/landing/Figure.astro
new file mode 100644
index 00000000..beff06ae
--- /dev/null
+++ b/website/src/components/landing/Figure.astro
@@ -0,0 +1,124 @@
+---
+// A hand-drawn figure: the image in a white frame (the drawings are on
+// white, so they sit on any band), a caption with a bold title, and a
+// "Text version" disclosure that holds the alt text, so every reader can get
+// what the drawing says. Blue in a figure means FORA's path, and nothing else.
+// The images are square PNGs under public/resources///.
+// Clicking a figure (or pressing Enter or Space on it) opens it full screen; a
+// click or Escape closes it. Ported from the zoom on mirin.pro
+// (cm-personal-blog, BaseHead.astro), with keyboard support added.
+//
+//
+interface Props {
+ src: string;
+ /** The text version: the image's alt text, also shown in the disclosure. */
+ alt: string;
+ /** Bold, first in the caption. */
+ title: string;
+ /** Follows the title; may hold inline HTML. */
+ caption?: string;
+ width?: number;
+ height?: number;
+}
+const { src, alt, title, caption, width = 1024, height = 1024 } = Astro.props;
+---
+
+
+ {title}{caption && <> >}
+
+ Text version
+
{alt}
+
+
+
+
+
+
+
+
diff --git a/website/src/components/landing/ForaGlance.astro b/website/src/components/landing/ForaGlance.astro
new file mode 100644
index 00000000..8798447d
--- /dev/null
+++ b/website/src/components/landing/ForaGlance.astro
@@ -0,0 +1,93 @@
+---
+// At a glance, FORA's column only: the hub's pair of cards, "FORA for your
+// revenue team" (Money) and "What FORA takes on your stack" (Your stack),
+// side by side. The rows are GLANCE_GROUPS' labels from
+// src/data/content-marketplaces.ts, the same labels the vendor pages'
+// GlanceCard uses, so the hub and the deep pages read the same. GlanceCard
+// itself always has a vendor column; this is the one-column form.
+//
+// Each card is a table, so screen readers get row headers. The cards sit side
+// by side and stack below 640px of their own width (a container query); a
+// card narrower than 420px puts each label above its value. The card titles
+// are the only blue. Cells hold HTML (statusTag(), cite(), links, ).
+//
+//
+interface Props {
+ cards: { title: string; group: string; rows: { label: string; cell: string }[] }[];
+}
+const { cards } = Astro.props;
+---
+
+
+ {cards.map((c) => (
+
+
+
+
{c.title}
+
+
+
{c.group}
+ {c.rows.map((r) => (
+
+
{r.label}
+
+
+ ))}
+
+
+
+ ))}
+
+
+
+
diff --git a/website/src/components/landing/GlanceCard.astro b/website/src/components/landing/GlanceCard.astro
new file mode 100644
index 00000000..59502769
--- /dev/null
+++ b/website/src/components/landing/GlanceCard.astro
@@ -0,0 +1,108 @@
+---
+// At a glance: the facts a reader acts on, as label and value pairs in
+// groups, with the vendor and FORA side by side. The FORA heading is the only
+// blue. A vendor page builds its groups with glanceGroups() from
+// src/data/content-marketplaces.ts, so every vendor page has the same rows in
+// the same order. A cell may hold inline HTML: a StatusTag (statusTag()), a
+// citation marker (cite()), a link, .
+//
+// It is a table, so screen readers get row and column headers. It sizes by
+// its own width (a container query), not the window: below 640px of card
+// width each row stacks into its label and two labelled values. Reads the
+// band's colour variables, so it works on any tone.
+//
+//
+interface Props {
+ vendor: string;
+ groups: { name: string; rows: { label: string; cells: readonly [string, string] }[] }[];
+ /** The table's accessible name. */
+ label?: string;
+}
+const { vendor, groups, label = 'At a glance' } = Astro.props;
+---
+
+
+
+
+
+
{vendor}
FORA
+
+
+ {groups.map((g) => (
+ <>
+
{g.name}
+ {g.rows.map((r) => (
+
+
{r.label}
+
+
+
+ ))}
+ >
+ ))}
+
+
+
+
+
+
diff --git a/website/src/components/landing/KeyFacts.astro b/website/src/components/landing/KeyFacts.astro
new file mode 100644
index 00000000..1094de4c
--- /dev/null
+++ b/website/src/components/landing/KeyFacts.astro
@@ -0,0 +1,61 @@
+---
+// Key facts: three or four numbers drawn large, each with a line on what it
+// means and, optionally, a source line. At most one strip per page, only
+// numbers the page already sources and that aren't in its GlanceCard. The
+// numbers are in the heading colour, not blue: a vendor's number isn't
+// FORA's. `text` and `source` may hold inline HTML, such as a citation marker
+// (cite()) or a link.
+//
+// Four across; two by two in a narrower column; one column on a phone (a
+// container query, so it follows its own width). Reads the band's variables.
+//
+//
+interface Props {
+ items: { value: string; text: string; source?: string }[];
+}
+const { items } = Astro.props;
+---
+
+
+ {items.map((it) => (
+
+
{it.value}
+
+ {it.source && }
+
+ ))}
+
+
+
+
diff --git a/website/src/components/landing/LinkRow.astro b/website/src/components/landing/LinkRow.astro
index e5595eb7..8527197f 100644
--- a/website/src/components/landing/LinkRow.astro
+++ b/website/src/components/landing/LinkRow.astro
@@ -1,16 +1,19 @@
---
-// A row of "read more" links at the end of a section.
+// A row of "read more" links at the end of a section. `stacked` sets them as
+// a short list, one per line (the series navigation).
interface Props {
links: { href: string; text: string }[];
+ stacked?: boolean;
}
-const { links } = Astro.props;
+const { links, stacked = false } = Astro.props;
---
-
diff --git a/website/src/components/landing/MarketMap.astro b/website/src/components/landing/MarketMap.astro
new file mode 100644
index 00000000..7ff362a4
--- /dev/null
+++ b/website/src/components/landing/MarketMap.astro
@@ -0,0 +1,200 @@
+---
+// The market map of the content-marketplaces hub (/resources/content-marketplaces/):
+// one table, every player in the same columns, grouped by category of
+// channel. FORA's row is pinned first and is the only blue. Column heads are
+// the reader's questions. Cells hold inline HTML: citation markers (cite()),
+// status tags (statusTag()), links.
+//
+// Wide, it is a table that runs a little wider than the band, up to 1320px,
+// and scrolls inside its own box if the window is narrower than the table's
+// minimum. Below 760px the table gives way to cards, one per player, so the
+// page never scrolls sideways: each card shows the first `phoneCols` columns
+// after the player (on the hub, "What's sold" and "Who pays today"), and the
+// rest fold behind a "More" disclosure on the player's line (, so it works without JS and
+// from the keyboard). Only one of the two is displayed at a time, so a screen
+// reader reads the map once. Reads the band's variables.
+//
+//
+interface Row {
+ player: string;
+ /** One HTML cell per column after the first. */
+ cells: string[];
+}
+interface Props {
+ /** Every column head, the player column first. */
+ columns: string[];
+ fora: Row;
+ groups: { name: string; rows: Row[] }[];
+ /** The table's accessible name. */
+ label: string;
+ /** On phones, how many columns after the player each card shows before "More". */
+ phoneCols?: number;
+}
+const { columns, fora, groups, label, phoneCols = 2 } = Astro.props;
+const heads = columns.slice(1);
+const fields = (r: Row) => r.cells.map((c, i) => ({ head: heads[i], html: c }));
+---
+
+ {r.cells.length > phoneCols && (
+
+ More on {r.player}
+
+ {fields(r).slice(phoneCols).map((f) =>
{f.head}
)}
+
+
+ )}
+
+ ))}
+
+ >
+ ))}
+
+
+
+
diff --git a/website/src/components/landing/MoneyTiles.astro b/website/src/components/landing/MoneyTiles.astro
new file mode 100644
index 00000000..c0e21a9d
--- /dev/null
+++ b/website/src/components/landing/MoneyTiles.astro
@@ -0,0 +1,79 @@
+---
+// How money reaches a publisher, one small tile per channel (the hub's hub-4):
+// who pays, who sits in the middle, who is paid, then three short lines. A
+// CSS grid of tiles rather than one drawing, so it reflows: three across,
+// then two, then one (a container query on its own width). The text is real
+// text, so it reads aloud and copies.
+//
+// Grey boxes and arrows for the other channels; the `fora` tile alone is
+// blue, because blue means FORA's path. The payee box carries a heavier
+// border. The first line of each tile is in the text colour, the rest muted.
+//
+//
+interface Props {
+ /** The group's accessible name. */
+ label: string;
+ tiles: { title: string; flow: string[]; lines: string[]; fora?: boolean }[];
+}
+const { label, tiles } = Astro.props;
+const ARROW = '';
+---
+
+
+
diff --git a/website/src/components/landing/PointList.astro b/website/src/components/landing/PointList.astro
index 0e0eb338..62a4bf8a 100644
--- a/website/src/components/landing/PointList.astro
+++ b/website/src/components/landing/PointList.astro
@@ -1,19 +1,55 @@
---
-// A checked list of points, each a bold lead-in and a sentence.
+// A checked list of points, each a bold lead-in and a sentence. The text may
+// hold inline HTML (a link, , a citation marker). `numbered` makes it a
+// checklist of tasks: numbered square boxes instead of ticks. Reads the
+// band's variables, so it works on any tone.
+//
+// The ticks are blue, because blue means FORA: use the default for FORA's own
+// points (what FORA does, what works today). `tone="neutral"` makes them grey
+// for points that are a vendor's or nobody's in particular: where a vendor is
+// ahead, a vendor's column, questions to ask. The numbered boxes are always
+// grey.
+//
+//
interface Props {
- items: { title: string; text: string }[];
+ /** A point may be text only, with no bold lead-in. */
+ items: { title?: string; text: string }[];
+ numbered?: boolean;
+ tone?: 'fora' | 'neutral';
}
-const { items } = Astro.props;
+const { items, numbered = false, tone = 'fora' } = Astro.props;
+const Tag = numbered ? 'ol' : 'ul';
---
-
+
{items.map((it) => (
-
{it.title} {it.text}
+
{it.title && <>{it.title} >}
))}
-
+
diff --git a/website/src/components/landing/PullQuote.astro b/website/src/components/landing/PullQuote.astro
index 43850cb6..7afea891 100644
--- a/website/src/components/landing/PullQuote.astro
+++ b/website/src/components/landing/PullQuote.astro
@@ -1,7 +1,10 @@
---
// A pull quote: one sentence quoted verbatim, who said it, and a link to the
// source. Used on /resources/transparent-collective/. The quote is plain text;
-// the curly quotation marks come from the stylesheet.
+// the curly quotation marks come from the stylesheet. `small` sets it at a
+// smaller size, for a quote inside a Callout. A quote is always someone
+// else's words, so its edge is grey (the band's label colour), not blue:
+// blue means FORA.
interface Props {
quote: string;
/** Who said it: a company, a regulator, a person. */
@@ -9,10 +12,11 @@ interface Props {
/** The source's title, linked to `href`. */
source: string;
href: string;
+ small?: boolean;
}
-const { quote, who, source, href } = Astro.props;
+const { quote, who, source, href, small = false } = Astro.props;
---
-
+
{quote}
{who}, {source}
@@ -21,7 +25,7 @@ const { quote, who, source, href } = Astro.props;
.pq {
margin: 2.5rem 0;
padding: 0.25rem 0 0.25rem 1.5rem;
- border-left: 3px solid var(--lp-link);
+ border-left: 3px solid var(--lp-label);
max-width: 46rem;
}
blockquote { margin: 0; }
@@ -38,4 +42,6 @@ const { quote, who, source, href } = Astro.props;
figcaption { font-size: 0.875rem; line-height: 1.5; color: var(--lp-label); }
figcaption::before { content: '\2014\00a0'; }
a { color: var(--lp-link); }
+ .small { margin: 0 0 1.5rem; padding-left: 1.125rem; }
+ .small p { font-size: clamp(1.0625rem, 1.8vw, 1.25rem); margin-bottom: 0.625rem; }
diff --git a/website/src/components/landing/ResourceTable.astro b/website/src/components/landing/ResourceTable.astro
new file mode 100644
index 00000000..36409082
--- /dev/null
+++ b/website/src/components/landing/ResourceTable.astro
@@ -0,0 +1,95 @@
+---
+// A resource page's one table, with columns of its own: a heading cell per
+// row, then plain cells. Cells hold inline HTML: links, , , status
+// tags (statusTag()) and citation markers (cite()). Used on
+// /resources/web-bot-auth/ ("Who signs"); StandardsMapTable is the same table
+// with the standards map's fixed columns.
+//
+// It sizes by its own width (a container query): below 640px each row becomes
+// a closed , its summary the row's heading and its first cell, so a
+// phone never scrolls sideways. With `labelAll` (rows that are questions, with
+// a column per answer) the summary is the heading alone and every cell shows
+// under its column's name. Reads the band's colour variables.
+//
+//
+interface Props {
+ /** Column headings; the first heads the row-heading column. */
+ cols: string[];
+ /** One array per row, as many cells as `cols`; the first is the row heading. */
+ rows: string[][];
+ /** The table's accessible name. */
+ label: string;
+ /** On phones, label every cell with its column, the first included. */
+ labelAll?: boolean;
+}
+const { cols, rows, label, labelAll = false } = Astro.props;
+---
+
+
+
diff --git a/website/src/components/landing/SectionDivider.astro b/website/src/components/landing/SectionDivider.astro
new file mode 100644
index 00000000..77168897
--- /dev/null
+++ b/website/src/components/landing/SectionDivider.astro
@@ -0,0 +1,29 @@
+---
+// A labelled rule, such as "For your legal and product teams": the detail
+// below it is optional reading. At most one per page. Reads the band's
+// variables.
+//
+//
+interface Props {
+ label: string;
+}
+const { label } = Astro.props;
+---
+
+ {label}
+
+
+
diff --git a/website/src/components/landing/SeriesNav.astro b/website/src/components/landing/SeriesNav.astro
new file mode 100644
index 00000000..40720bd3
--- /dev/null
+++ b/website/src/components/landing/SeriesNav.astro
@@ -0,0 +1,32 @@
+---
+// "More in this series": the other pages of the content-marketplaces series,
+// in the series' order, with the current page left out. The same block ends
+// every page of the series, before the closing call. The list is SERIES in
+// src/data/content-marketplaces.ts. Reads the band's variables.
+//
+//
+import LinkRow from './LinkRow.astro';
+import { SERIES } from '../../data/content-marketplaces';
+
+interface Props {
+ /** The href of the page this block sits on. */
+ current: string;
+}
+const { current } = Astro.props;
+const links = SERIES.filter((p) => p.href !== current).map((p) => ({ href: p.href, text: p.title }));
+---
+
+
+
diff --git a/website/src/components/landing/SourceList.astro b/website/src/components/landing/SourceList.astro
index 93d34b27..1f769cab 100644
--- a/website/src/components/landing/SourceList.astro
+++ b/website/src/components/landing/SourceList.astro
@@ -1,6 +1,11 @@
---
// A page's sources, in groups: each entry's title (linked), publisher and URL. A source with no public URL gives `where` instead.
-// Used on /resources/transparent-collective/.
+// Used on /resources/transparent-collective/ and /resources/music-dsp/.
+//
+// `numbered` numbers the entries continuously across groups, shows each
+// number, and gives each entry id="src-N", so a marker cite(N) in the copy
+// (src/lib/resource-markup.ts) links to it. Number the sources in order of
+// first citation on the page.
interface Source {
title: string;
publisher: string;
@@ -14,17 +19,21 @@ interface Props {
groups: { name: string; items: Source[] }[];
/** Titles and publishers only; the URL line shows only for a source with no link. */
compact?: boolean;
+ /** Number the entries across groups, each with id="src-N". */
+ numbered?: boolean;
}
-const { groups, compact = false } = Astro.props;
+const { groups, compact = false, numbered = false } = Astro.props;
+// The number of each group's first entry.
+const starts = groups.map((_, i) => groups.slice(0, i).reduce((n, g) => n + g.items.length, 1));
const shown = (href: string) => (href.startsWith('/') ? `fora-protocol.org${href}` : href.replace(/^https?:\/\//, ''));
---
-
- {groups.map((g) => (
+
+ {groups.map((g, gi) => (
{g.name}
-
- {g.items.map((s) => (
-
+
+ {g.items.map((s, i) => (
+
{s.href ? {s.title} : {s.title}}
{s.publisher}
{(!compact || !s.href) && {s.href ? shown(s.href) : s.where}}
@@ -68,4 +77,16 @@ const shown = (href: string) => (href.startsWith('/') ? `fora-protocol.org${href
.compact a, .compact li > b { font-size: 0.9375rem; }
.compact .src-meta { font-size: 0.8125rem; color: var(--lp-label); }
.src-url { font-family: var(--mono); font-size: 0.75rem; color: var(--lp-label); overflow-wrap: anywhere; }
+ .numbered li { position: relative; padding-left: 2.25rem; scroll-margin-top: 5rem; }
+ .numbered li::before {
+ content: attr(data-n);
+ position: absolute;
+ left: 0;
+ top: 0.1rem;
+ font-family: var(--mono);
+ font-size: 0.8125rem;
+ color: var(--lp-label);
+ font-variant-numeric: tabular-nums;
+ }
+ .numbered li:target { outline: 2px solid var(--lp-link); outline-offset: 0.375rem; border-radius: 2px; }
diff --git a/website/src/components/landing/Split.astro b/website/src/components/landing/Split.astro
new file mode 100644
index 00000000..3065b4b7
--- /dev/null
+++ b/website/src/components/landing/Split.astro
@@ -0,0 +1,26 @@
+---
+// Two columns inside a band, below its header: the default slot on the left,
+// the `aside` slot on the right (a figure beside a callout, two columns of
+// prose). Stacks below 900px, the aside under the main column. Band's own
+// `aside` slot puts the band's header in the left column; use Split when the
+// header should run full width above both.
+//
+//
+interface Props {
+ /** Width of the right column. */
+ aside?: string;
+}
+const { aside = '1fr' } = Astro.props;
+---
+
+
+
+
+
+
diff --git a/website/src/components/landing/StandardsMapTable.astro b/website/src/components/landing/StandardsMapTable.astro
new file mode 100644
index 00000000..862677a2
--- /dev/null
+++ b/website/src/components/landing/StandardsMapTable.astro
@@ -0,0 +1,98 @@
+---
+// The standards map's table (/resources/standards-map/): one row per
+// standard, with its job, its status and who stands behind it, and who uses
+// it. Cells hold inline HTML: links, , for the job's name, citation
+// markers (cite()). The FORA row's name is FORA's blue, the one colour that
+// means FORA.
+//
+// It sizes by its own width (a container query): below 640px each row becomes
+// a closed , its summary the standard and its job, so a phone never
+// scrolls sideways. Reads the band's colour variables.
+//
+//
+interface Row {
+ name: string;
+ href?: string;
+ job: string;
+ status: string;
+ users: string;
+ /** FORA's own row. */
+ fora?: boolean;
+}
+interface Props {
+ rows: Row[];
+ /** The table's accessible name. */
+ label?: string;
+}
+const { rows, label = 'The standards map' } = Astro.props;
+const COLS = ['Standard', 'Its job, and what it answers', 'Status, and who stands behind it', 'Who uses it'];
+const name = (r: Row) => (r.href ? `${r.name}` : r.name);
+---
+
+
+
{COLS.map((c) =>
{c}
)}
+
+ {rows.map((r) => (
+
+
+
+
+
+
+ ))}
+
+
+
+ {rows.map((r) => (
+
+
+
+
{COLS[2]}
+
{COLS[3]}
+
+
+ ))}
+
+
+
+
diff --git a/website/src/components/landing/StatusTag.astro b/website/src/components/landing/StatusTag.astro
new file mode 100644
index 00000000..e68deb16
--- /dev/null
+++ b/website/src/components/landing/StatusTag.astro
@@ -0,0 +1,21 @@
+---
+// A status tag: how real a thing is (Live, Beta, Closed beta, Pilot, Preview,
+// Experimental, Design, Roadmap, Not public; the meanings are in
+// src/data/content-marketplaces.ts).
+// Used inline, in cards, in table cells and beside headings. Every design or
+// roadmap item carries one, so nothing designed reads as built.
+//
+// The markup comes from statusTag() in src/lib/resource-markup.ts, which an
+// HTML string (a GlanceCard cell) calls directly; the styles are global, in
+// styles/landing.css, so both look the same. Reads the band's variables.
+//
+//
+import { statusTag } from '../../lib/resource-markup';
+import type { StatusKey } from '../../data/content-marketplaces';
+
+interface Props {
+ status: StatusKey;
+}
+const { status } = Astro.props;
+---
+
diff --git a/website/src/components/landing/StepList.astro b/website/src/components/landing/StepList.astro
new file mode 100644
index 00000000..6a357b04
--- /dev/null
+++ b/website/src/components/landing/StepList.astro
@@ -0,0 +1,48 @@
+---
+// Numbered steps in mono rings: a vendor's process, or the order of rules on
+// your CDN. The rings are grey, because blue means FORA; `fora` turns them
+// blue for FORA's own steps. Each step is a bold lead-in and HTML text (links,
+// , citation markers); a step may be text only. Reads the band's
+// variables, so it works on any tone and inside a Callout.
+//
+//
+interface Props {
+ items: { title?: string; text: string }[];
+ fora?: boolean;
+}
+const { items, fora = false } = Astro.props;
+---
+
+ {items.map((it) => (
+
{it.title && <>{it.title} >}
+ ))}
+
+
+
diff --git a/website/src/consts.ts b/website/src/consts.ts
index 204e5d86..5a2b94de 100644
--- a/website/src/consts.ts
+++ b/website/src/consts.ts
@@ -70,6 +70,7 @@ export const LANDING_PAGES = [
{ key: 'publishers', href: '/publishers/', label: 'For publishers' },
{ key: 'music', href: '/music/', label: 'For music' },
{ key: 'resources', href: '/resources/', label: 'Resources' },
+ { key: 'about', href: '/about/', label: 'About' },
] as const;
export type LandingPageKey = (typeof LANDING_PAGES)[number]['key'];
diff --git a/website/src/content/docs/architecture/deployment-models.mdx b/website/src/content/docs/architecture/deployment-models.mdx
index 71a94a7c..99075ba8 100644
--- a/website/src/content/docs/architecture/deployment-models.mdx
+++ b/website/src/content/docs/architecture/deployment-models.mdx
@@ -30,7 +30,7 @@ The Exchange validates incoming `CatalogService.PushResources` requests against
## Model 1: Exchange-Managed (Recommended, Enterprise)
-The Exchange operator deploys and maintains the FORA edge function on the provider's CDN on their behalf. This is the **Prebid Server model** -- "we handle everything."
+The Exchange operator deploys and maintains the FORA edge function on the provider's CDN on their behalf.
### How It Works
diff --git a/website/src/content/docs/architecture/nfr.mdx b/website/src/content/docs/architecture/nfr.mdx
index 7dc55efe..f0a5a216 100644
--- a/website/src/content/docs/architecture/nfr.mdx
+++ b/website/src/content/docs/architecture/nfr.mdx
@@ -1,6 +1,6 @@
---
title: "Non-Functional Requirements"
-description: "FORA performance, latency, throughput, durability, and cost targets modeled on ad-tech scale"
+description: "FORA performance, latency, throughput, durability and cost targets"
---
## Context: The Numbers That Drive Everything
@@ -16,15 +16,6 @@ FORA targets the **inference/RAG segment** -- real-time, per-request resource ac
**Inference traffic estimate**: 20% of 50B = ~10B requests/day = ~115K requests/second globally. A single large Exchange might handle 1-10% of this = **1K-10K RPS**.
-### Ad-Tech Benchmarks
-
-- OpenRTB exchanges: **100ms total latency budget**, broadcast to 50+ DSPs
-- Individual SSP bid request assembly: **2-5ms**
-- AWS RTB Fabric: sub-millisecond container-to-container communication
-- Production exchanges handle **millions of QPS**
-
-FORA doesn't need ad-tech scale day one, but the architecture must not preclude it.
-
## Latency Requirements
### Per-Hop Latency Budgets
@@ -47,7 +38,7 @@ Total agent-perceived latency target: **< 200ms end-to-end** for the transaction
|---|---|---|
| Agent → Broker: DiscoveryRequest | **2ms** | **5ms** |
| Broker → N Exchanges (parallel) | **15ms** | **60ms** |
-| Broker: Select best offer | **1ms** | **3ms** |
+| Broker: Order offers | **1ms** | **3ms** |
| Broker → Exchange: ExecuteTransaction | **20ms** | **100ms** |
| Total Broker overhead | **< 40ms** | **< 170ms** |
@@ -113,7 +104,7 @@ Total agent-perceived latency target: **< 200ms end-to-end** for the transaction
| Exchange down | Agents can't discover or transact | Broker tries other Exchanges |
| Billing Adapter down | Transactions fail | Policy decision: reject or defer |
| Edge Function down | Bots aren't blocked | CDN continues serving; edge function is defense-in-depth |
-| Broker down | Agents lose multi-Exchange comparison | Agents query Exchanges directly |
+| Broker down | Agents lose multi-Exchange discovery | Agents query Exchanges directly |
| Transaction log store down | **Critical** | Exchange must stop issuing signed URLs |
## Cost Model
diff --git a/website/src/content/docs/architecture/production-architecture.mdx b/website/src/content/docs/architecture/production-architecture.mdx
index 1385b8ba..19b596b8 100644
--- a/website/src/content/docs/architecture/production-architecture.mdx
+++ b/website/src/content/docs/architecture/production-architecture.mdx
@@ -5,7 +5,7 @@ description: "FORA production architecture: components, roles, protocol flows, a
## What We're Building
-Open-source infrastructure components for metered resource access. Like Prebid.js -- the protocol and reference implementations are free; wiring them into an Exchange operator's billing or a provider's CMS is custom consulting work.
+Open-source infrastructure components for metered resource access. The protocol and FORA Server are free; an Exchange operator wires them into its own billing, and a provider into its own CMS.
### Components
@@ -17,19 +17,6 @@ Open-source infrastructure components for metered resource access. Like Prebid.j
| **Edge Function** | Provider | Provider's CDN (CloudFront/Cloudflare/Fastly) |
| **Client SDK** | AI agent developers | Low-level generated Connect client for direct Exchange RPC access (no orchestration logic) |
-### Role Mapping (Ad-Tech Analogy)
-
-```
-Ad Tech FORA / CoMP PoC Equivalent
-─────────────────────────────────────────────────────────────────────
-Browser (user) → AI Agent → MCP Tool
-DSP SDK (prebid.js) → Agent SDK (embedded) → (not built yet)
-DSP server → FORA Broker → (not built yet)
-SSP / Exchange → Exchange → Gateway + Registry
-Provider → Resource Provider → demo.fora-protocol.org
-CDN / Ad Creative → Resource Delivery Endpoint → CloudFront + S3
-```
-
## Protocol Architecture
### Core Principle: Exchange Protocol Is the Standard
@@ -38,7 +25,7 @@ The Exchange interface is the core protocol. Both agents and Brokers are clients
```
With Broker (multi-exchange, budget control, reporting):
- Agent → Broker → Exchange A, B, C → select best → deliver
+ Agent → Broker → Exchange A, B, C → offers in order of relevance → agent buys → deliver
Without Broker (direct, single exchange):
Agent → Exchange → deliver
diff --git a/website/src/content/docs/brand.mdx b/website/src/content/docs/brand.mdx
index cf5b4265..9ce0c3dc 100644
--- a/website/src/content/docs/brand.mdx
+++ b/website/src/content/docs/brand.mdx
@@ -88,7 +88,8 @@ the agent receives, instead of adjectives. No "revolutionary", "seamless",
## Logo
-The mark is a wax seal pressed with the open forum. Before signatures and
+FORA stands for Federated Open Resource Access, and *fora* is also the plural of
+*forum*. That's why the mark lays the plan of a Roman forum over a wax seal. Before signatures and
cryptography, a wax seal proved who sent a letter and that nobody had opened it
on the way, and any recipient could check it without asking a central office.
That is what FORA's signed offers and acceptances do. The impression is an open
diff --git a/website/src/content/docs/components/agent-sdk/fetch-flow.mdx b/website/src/content/docs/components/agent-sdk/fetch-flow.mdx
index 5fba06b7..49c14b9d 100644
--- a/website/src/content/docs/components/agent-sdk/fetch-flow.mdx
+++ b/website/src/content/docs/components/agent-sdk/fetch-flow.mdx
@@ -45,15 +45,17 @@ client.Fetch(ctx, "https://techcrunch.com/premium/article.html")
| - Collect ResourceResponses (respect deadline, drop slow responders)
| - Flatten all Offers into candidate list
|
- | 5. Select best offer
+ | 5. Choose an offer (the agent's own policy)
| SelectionEngine.Select(candidates, constraints)
| - Choose the Offer whose term's restrictions
| (function, geo, user-type) the agent can honour
| - Group by ResourceIdentity (deduplication)
- | - Rank: subscription offers first (rate=0)
+ | - Apply the agent's selection policy, for example:
+ | subscription offers first (rate=0)
| -> preferred Exchange offers second
| -> lowest unit_cost third
- | - Return winning Offer + its Exchange
+ | The SDK runs inside the agent, so this choice is the agent's
+ | - Return the chosen Offer + its Exchange
|
| 6. Budget check (exact amount)
| BudgetTracker.Check(offer.Pricing.Rate)
@@ -65,7 +67,7 @@ client.Fetch(ctx, "https://techcrunch.com/premium/article.html")
| TransactionExecutor.Execute(ctx, exchange, offer)
| - Build TransactionRequest embedding the full signed offer (reflected back)
| - Authenticate via RFC 9421 HTTP Message Signature in the HTTP headers
- | - Send ExecuteTransaction RPC to winning Exchange
+ | - Send ExecuteTransaction RPC to the chosen offer's Exchange
| - On DenialReason: return typed error (see Error Handling below)
| - On success: extract signed URL from retrieval_endpoint, as issued.
| The SDK does not verify it; the delivery edge does, at step 9
@@ -118,7 +120,7 @@ client.Fetch(ctx, "https://techcrunch.com/premium/article.html")
| Resolve Exchanges (cache miss) | 50ms | 200ms | HTTP fetch of fora.json |
| Pre-flight budget check | <1ms | <1ms | In-memory arithmetic |
| DiscoverResources RPC | 10ms | 50ms | Per NFR targets |
-| Select best offer | <1ms | <1ms | In-memory sort |
+| Choose an offer | <1ms | <1ms | In-memory sort |
| Budget check (exact) | <1ms | <1ms | In-memory arithmetic |
| ExecuteTransaction RPC | 20ms | 100ms | Per NFR targets |
| Record spend | <1ms | <1ms | In-memory write |
@@ -148,9 +150,9 @@ client.FetchBatch(ctx, [url1, url2, url3])
| - Exchange returns OfferGroups (one per URI)
| - If an Exchange covers multiple domains, one query per domain
|
- | 5. Select best offer per URI
- | For each URI: merge offers from all Exchanges, select best
- | - Same ranking: subscription > preferred > lowest unit_cost
+ | 5. Choose an offer per URI (the agent's own policy)
+ | For each URI: merge offers from all Exchanges, choose one
+ | - Same agent policy as for a single URL
| - Deduplicate by ResourceIdentity across Exchanges
|
| 6. Budget check (total)
@@ -161,7 +163,7 @@ client.FetchBatch(ctx, [url1, url2, url3])
| - Individual MaxPerRequest checked per URI
|
| 7. Execute transactions (batch, grouped by Exchange)
- | Group selected offers by winning Exchange
+ | Group the chosen offers by Exchange
| For each Exchange: send ONE batch TransactionRequest with items[]
| - Each item: the full signed offer (reflected back)
| - Exchange returns TransactionResultItem per offer
@@ -186,7 +188,7 @@ client.FetchBatch(ctx, [url1, url2, url3])
### Batch Optimizations
- **One ResourceQuery per Exchange**: N URLs in one request instead of N separate requests. Reduces network round trips from N*M to M (where M = number of Exchanges).
-- **One TransactionRequest per winning Exchange**: batch items[] instead of individual transactions.
+- **One TransactionRequest per Exchange**: batch items[] instead of individual transactions.
- **Parallel content fetch**: all signed URLs fetched concurrently after all transactions complete.
- **Non-atomic**: individual URLs can fail while others succeed. Each FetchResult carries its own Err.
diff --git a/website/src/content/docs/components/agent-sdk/overview.mdx b/website/src/content/docs/components/agent-sdk/overview.mdx
index 7a4d918a..0b408d05 100644
--- a/website/src/content/docs/components/agent-sdk/overview.mdx
+++ b/website/src/content/docs/components/agent-sdk/overview.mdx
@@ -383,7 +383,8 @@ Agent SDK (planned layer)
| Parallel fanout to multiple Exchanges.
|
+-- SelectionEngine
- | Ranks offers: subscription > preferred MP > lowest unit_cost.
+ | Applies the agent's own selection policy, for example:
+ | subscription > preferred MP > lowest unit_cost.
| Deduplicates by ResourceIdentity across Exchanges.
| Chooses the Offer whose term the agent can honour (one offer per term).
|
@@ -394,7 +395,7 @@ Agent SDK (planned layer)
| Checks budget BEFORE ExecuteTransaction, not after.
|
+-- TransactionExecutor
- | Sends ExecuteTransaction RPC to winning Exchange.
+ | Sends ExecuteTransaction RPC to the chosen offer's Exchange.
| Stateless offer verification via exchange_signature.
| Idempotency key on every request.
|
@@ -419,7 +420,7 @@ Agent SDK (planned layer)
The SDK reuses the same Go interfaces defined for the standalone Broker:
- `ExchangeRegistry` interface is identical
-- `SelectionEngine` interface is identical (same `SelectionPolicy` abstraction)
+- `SelectionEngine` interface is identical (same `SelectionPolicy` abstraction); the SDK plugs in the agent's own policy
- `BudgetTracker` interface is identical
The concrete implementations differ in persistence strategy (in-memory vs Redis), but the interfaces are shared. The SDK also has a `UsageReporter`, which files the agent's usage reports with each issuing Exchange. The standalone Broker has no counterpart, because an agent always files its own reports, for a purchase made through a Broker too.
@@ -463,7 +464,7 @@ The concrete implementations differ in persistence strategy (in-memory vs Redis)
| Exchange trust | All configured = verified | Graduated trust model |
| Process lifecycle | Tied to agent process | Independent process |
-**Same logic, different packaging.** The selection algorithm, unit_cost comparison, ResourceIdentity deduplication, subscription preference ranking, and budget arithmetic are identical.
+**Same logic, different packaging, one difference in who chooses.** ResourceIdentity deduplication and budget arithmetic are identical. The SDK runs inside the agent, so it applies the agent's own selection policy and buys the offer it chooses. The Broker returns the offers in its order of relevance, and the agent decides.
## Language Bindings
diff --git a/website/src/content/docs/components/broker/budget-enforcement.mdx b/website/src/content/docs/components/broker/budget-enforcement.mdx
deleted file mode 100644
index 836300e6..00000000
--- a/website/src/content/docs/components/broker/budget-enforcement.mdx
+++ /dev/null
@@ -1,177 +0,0 @@
----
-title: "Budget Enforcement"
-description: "How the Broker tracks and enforces per-request, per-session, and per-period budgets to prevent overspend across AI resource transactions."
----
-
-## Overview
-
-The Broker applies three budget layers, analogous to how DSPs manage advertiser budgets in OpenRTB. The budget is a **filter over offers** at `BrokerService.Resolve`: the Broker drops every offer that would take the agent past its budget, and answers a URI left with no offer with `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED`. The Exchange's balance check stays the authoritative gate on what is actually charged.
-
-How a budget is computed — which price an offer counts at and how spend is counted — is the Broker's own behaviour, and the protocol does not specify it. This page describes one design. What the protocol does fix is how a budget's effect is reported, and that Resolve never charges:
-
-- **Resolve never charges.** Returning an offer reserves nothing.
-- **A budget is reported only as `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED`**: never as `OFFER_ABSENCE_REASON_NOT_AUTHORIZED`, and never as a non-OK error such as `RESOURCE_EXHAUSTED`.
-
-## Budget Layers
-
-```mermaid
-graph TB
- subgraph "Budget Enforcement Layers"
- L1[Layer 1: Per-Offer RequestConstraints.max_price offers above it are filtered]
- L2[Layer 2: Per-Session spend tracking in-memory, reset on restart]
- L3[Layer 3: Per-Period per-month, per-day persisted in Redis]
- L4[Layer 4: Per-Exchange balance awareness advisory, not enforced]
- end
-
- L1 --- L2
- L2 --- L3
- L3 --- L4
-```
-
-| Layer | Source of Truth | Enforcement | Crash Behavior | Storage |
-|---|---|---|---|---|
-| **Per-offer** | `RequestConstraints.max_price` | Drop an offer priced above it | Stateless -- always applied | None |
-| **Per-session** | Broker in-memory map | Drop an offer that would take the session's spend past the session budget | Lost on crash; agent re-declares | In-memory |
-| **Per-period** | Redis, keyed by `budget_scope`, else the verified agent identity | Drop an offer that would take the period's spend past `period_budget` | Survives restart (persisted in Redis) | Redis |
-| **Per-Exchange** | Exchange's billing adapter | Advisory -- Exchange ultimately decides | Not Broker's state | External |
-
-## Budget Scope Identifiers
-
-`RequestConstraints` supports a `budget_scope` identifier so the agent can say "track this transaction against user X's monthly budget":
-
-```go
-type RequestConstraints struct {
- // ...existing fields...
- BudgetScope string // optional: "user:alice", "team:engineering", "project:q4-research"
- PreferredExchanges []string // exchange domains with existing subscriptions/relationships
- PeriodBudget *Cost // per-period budget limit (proto: period_budget)
- BudgetPeriod *time.Duration // budget period, e.g. 720h = 30 days (proto: budget_period)
-}
-```
-
-Per-period budgets are keyed by `budget_scope` (or the verified agent identity if no scope is provided) and persisted in Redis with TTLs matching the period (e.g., monthly keys expire at month-end).
-
-## Session Budget Tracker
-
-```go
-type BudgetTracker interface {
- // SetSessionBudget declares the budget for a session.
- SetSessionBudget(sessionID string, amount float64, currency string)
-
- // CanAfford reports whether buying an offer at this price would keep the
- // spend within the budget. It reserves nothing.
- CanAfford(amount float64, currency string) bool
-
- // Record adds a purchase's charged amount to the spend.
- Record(ctx context.Context, cost *forav1.Cost)
-
- // Remaining returns the remaining budget for the session.
- Remaining(sessionID string) (float64, string)
-}
-
-type inMemoryBudget struct {
- mu sync.Mutex
- sessions map[string]*sessionBudget
-}
-
-type sessionBudget struct {
- limit float64
- currency string
- spent float64
- txnCount int
- lastTxn time.Time
-}
-
-func (b *inMemoryBudget) CanAfford(amount float64, currency string) bool {
- b.mu.Lock()
- defer b.mu.Unlock()
- // A budget counts amounts in its own currency. This version of the protocol
- // defines no currency conversion.
- return (b.current.spent + amount) <= b.current.limit
-}
-```
-
-## Batch Selection and Recording Spend
-
-For batch (multi-URI) queries, the selection engine sums the prices of the offers it selects. If the total exceeds the remaining budget, it drops lower-priority URIs until the total fits (see the [Selection Engine](/components/broker/selection-engine/) batch pipeline). This is still filtering: nothing is reserved by `Resolve`.
-
-In this design, spend is recorded when a purchase completes: for a purchase through the Broker, the Broker adds each entry of `BrokerTransactionResponse.totals` (one `Cost` per currency, covering only the charged items) to the budget of the same currency.
-
-```go
-func (b *inMemoryBudget) RecordPurchase(ctx context.Context, totals []*forav1.Cost) {
- b.mu.Lock()
- defer b.mu.Unlock()
- for _, t := range totals {
- if t.Currency != b.current.currency {
- continue // amounts are never converted between currencies
- }
- b.current.spent += t.Amount
- }
- b.current.txnCount++
- b.current.lastTxn = time.Now()
-}
-```
-
-## Overspend Prevention
-
-The Broker applies the budget while it selects offers, before it returns them. An offer the agent cannot afford is never returned, so the agent never holds it to buy.
-
-### Budget Exhaustion Mid-Session
-
-When the session or period budget leaves no affordable offer for a URI:
-
-1. `Resolve` still returns OK. That URI's `OfferGroup` carries no offers and `absence_reason: OFFER_ABSENCE_REASON_BUDGET_EXCEEDED`; when no URI has an affordable offer, `DiscoveryResponse.absence_reason` is `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED`.
-2. It is not an error. The Broker never answers with `RESOURCE_EXHAUSTED` or `NOT_AUTHORIZED` for a budget.
-3. The agent can re-declare a higher budget, or buy directly from an Exchange without the Broker.
-
-```go
-price := budgetPrice(offer.Pricing) // the Broker's own choice of price
-if !budget.CanAfford(price, offer.Pricing.Currency) {
- // Drop the offer. If none remains for this URI, answer it with
- // OFFER_ABSENCE_REASON_BUDGET_EXCEEDED. Nothing is charged.
- continue
-}
-```
-
-## Ephemeral vs. Persisted Budget State
-
-Session budget state is ephemeral. If the Broker crashes, the agent re-declares its budget on reconnect. Per-period budgets (monthly, daily) are persisted in Redis and survive restarts.
-
-This is a deliberate design choice: the Exchange's billing adapter is the authoritative source of truth for "can this requester afford this transaction," not the Broker.
-
-## Currencies
-
-Currency conversion is out of scope for this version of the protocol. A budget is stated in one currency (`period_budget.currency`), and only amounts in that currency count against it. Purchase totals are never summed across currencies: `BrokerTransactionResponse.totals` carries one entry per currency, and `TransactionResponse.total_cost` is unset when the purchased items span more than one currency.
-
-## Budget Consistency at Scale
-
-The Broker's budget is a filter over the offers it returns, not a reservation. The Exchange is the authoritative point for what is charged, and denies an item, in the body, if the requester's balance is insufficient. This means:
-
-- If two Broker instances both return an offer the budget still allows, and the agent buys both, the spend can pass the budget by the price of the second purchase. Each purchase's totals are recorded when it completes, and later resolves filter against the new spend.
-- The Exchange's balance check still bounds what the requester can be charged.
-- The Broker's budget filter keeps the agent from buying what its budget does not allow; it is not a guarantee across concurrent purchases.
-
-This model is deliberately simple. Distributed budget consensus (e.g., two-phase commit across Broker instances) adds latency and complexity without value, because the Exchange already provides the authoritative enforcement.
-
-## Budget Reporting Metrics
-
-```go
-var (
- // Budget utilization per session
- budgetUtilization = promauto.NewHistogram(
- prometheus.HistogramOpts{
- Name: "broker_budget_utilization_ratio",
- Help: "Fraction of session budget spent (0.0 to 1.0+)",
- Buckets: []float64{0.0, 0.1, 0.25, 0.5, 0.75, 0.9, 1.0},
- },
- )
-
- // Budget exhaustion events
- budgetExhausted = promauto.NewCounter(
- prometheus.CounterOpts{
- Name: "broker_budget_exhausted_total",
- Help: "Number of URIs answered with BUDGET_EXCEEDED because the budget left no offer",
- },
- )
-)
-```
diff --git a/website/src/content/docs/components/broker/content-dedup.mdx b/website/src/content/docs/components/broker/content-dedup.mdx
deleted file mode 100644
index 46db844f..00000000
--- a/website/src/content/docs/components/broker/content-dedup.mdx
+++ /dev/null
@@ -1,164 +0,0 @@
----
-title: "Content Deduplication"
-description: "How the Broker deduplicates identical content offered by multiple Exchanges using ResourceIdentity matching, SimHash comparison, and canonical URL grouping."
----
-
-## Purpose
-
-The same resource can appear on multiple Exchanges. An AP wire story might be available from Exchange A and Exchange B. A syndicated article might be offered by the original provider's Exchange and a resource aggregator. The Broker deduplicates these offers so the agent pays the lowest price for each unique resource.
-
-## ResourceIdentity Matching
-
-Each Offer carries an optional `ResourceIdentity` that the Broker uses for deduplication. The identity key is derived from a priority-ordered set of fields:
-
-```go
-// contentIdentityKey produces a deduplication key from layered identity fields.
-// Matching priority: iptc_guid > doi > content_hash > canonical_url
-func contentIdentityKey(id *forav1.ResourceIdentity) string {
- if id == nil {
- return uuid.NewString() // no identity = unique (no dedup)
- }
- if id.IptcGuid != nil && *id.IptcGuid != "" {
- return "iptc:" + *id.IptcGuid
- }
- if id.Doi != nil && *id.Doi != "" {
- return "doi:" + *id.Doi
- }
- if id.ContentHash != nil && *id.ContentHash != "" {
- return "hash:" + *id.ContentHash
- }
- if id.CanonicalUrl != nil && *id.CanonicalUrl != "" {
- return "url:" + *id.CanonicalUrl
- }
- return uuid.NewString()
-}
-```
-
-### Identity Field Priority
-
-| Priority | Field | Description | Example |
-|---|---|---|---|
-| 1 (highest) | `iptc_guid` | IPTC globally unique identifier for news content | `urn:newsml:apnews.com:20260315:article-123` |
-| 2 | `doi` | Digital Object Identifier for academic/scientific content | `10.1038/s41586-026-01234-5` |
-| 3 | `content_hash` | Content fingerprint (method depends on hash_method: simhash-v1 for fuzzy dedup, sha256 for exact match) | `sha256:abc123...` |
-| 4 (lowest) | `canonical_url` | Canonical URL of the content | `https://apnews.com/article/ai-regulation` |
-
-When no `ResourceIdentity` is present on an offer, the offer is treated as unique and is never deduplicated. This is the conservative default -- it is better to pay twice for the same content than to accidentally merge two different articles.
-
-## SimHash Comparison
-
-For content available from multiple sources that may have minor editorial differences (e.g., AP wire stories republished with small modifications), strict hash matching (`content_hash`) may miss duplicates. The Broker supports SimHash comparison for fuzzy matching:
-
-- **Level 2 verification** uses SHA-256 exact hash -- content is byte-for-byte identical
-- **Level 1 verification** uses SimHash similarity -- content is structurally similar (e.g., 95%+ similarity)
-- **Level 0** -- no content verification
-
-SimHash comparison is applied when:
-
-1. Two offers share the same `canonical_url` but have different `content_hash` values
-2. Two offers from different Exchanges reference the same article by title/author metadata but have no shared identifier
-
-The SimHash comparison operates on the content identity metadata provided by the Exchange, not on the content itself (which is not yet fetched at selection time).
-
-## Canonical URL Grouping
-
-When higher-priority identifiers are not available, the Broker falls back to `canonical_url` grouping. This catches the common case where the same article is offered by multiple Exchanges that all reference the same provider URL.
-
-Canonical URL matching is exact string comparison after normalization:
-
-- Strip trailing slashes
-- Lowercase the hostname
-- Sort query parameters
-- Remove tracking parameters (utm_*, ref, etc.)
-
-## Cross-Exchange Dedup Strategy
-
-The full deduplication pipeline operates across all Exchange responses:
-
-```go
-type contentGroup struct {
- Key string
- Offers []rankedOffer
-}
-
-func deduplicateOffers(offers []rankedOffer) []rankedOffer {
- groups := make(map[string][]rankedOffer)
-
- for _, o := range offers {
- key := contentIdentityKey(o.Offer.Identity)
- groups[key] = append(groups[key], o)
- }
-
- var deduped []rankedOffer
- for _, group := range groups {
- // Sort by unit_cost ascending within group
- sort.Slice(group, func(i, j int) bool {
- return unitCostOf(group[i]) < unitCostOf(group[j])
- })
- deduped = append(deduped, group[0]) // cheapest wins
- }
- return deduped
-}
-```
-
-### How It Works
-
-1. **Collect offers** from all Exchange responses.
-2. **Compute identity key** for each offer using the priority-ordered fields.
-3. **Group offers** by identity key. Offers with the same key are considered duplicates.
-4. **Within each group**, sort by effective cost per unit (unit_cost) ascending.
-5. **Keep only the cheapest** offer from each group.
-
-### Example
-
-An agent requests content at `https://apnews.com/article/ai-regulation`. Two Exchanges respond:
-
-| Exchange | Offer ID | unit_cost | ResourceIdentity |
-|---|---|---|---|
-| Exchange Alpha | offer-a1 | $0.0003/token | `iptc:urn:newsml:apnews.com:20260315:ai-reg` |
-| Exchange Beta | offer-b1 | $0.0005/token | `iptc:urn:newsml:apnews.com:20260315:ai-reg` |
-
-Both offers share the same `iptc_guid`. The deduplicator groups them and selects `offer-a1` (lower unit_cost). The agent pays $0.0003/unit instead of $0.0005/unit.
-
-## Deduplication in Batch Queries
-
-For batch (multi-URI) queries, deduplication runs per URI. Each URI's offers are deduplicated independently:
-
-1. For each URI in the batch, collect all offers from all Exchange responses that match that URI.
-2. Within each URI's offer set, run the standard deduplication pipeline.
-3. Select the best (cheapest) offer per URI after dedup.
-
-Cross-URI deduplication (e.g., "the agent requested two URLs that happen to be the same article") is handled by the batch selection engine, which checks `ResourceIdentity` across URIs before executing transactions.
-
-## Deduplication Metrics
-
-The Broker tracks deduplication effectiveness through Prometheus metrics:
-
-```go
-var (
- // How often deduplication found the same content from multiple Exchanges
- deduplicationRate = promauto.NewHistogram(
- prometheus.HistogramOpts{
- Name: "broker_dedup_sets",
- Help: "Number of content groups with offers from multiple Exchanges",
- Buckets: []float64{0, 1, 2, 3, 5, 10},
- },
- )
-)
-```
-
-The `SelectionRationale` returned with every selection decision includes `DeduplicatedSets` -- the count of content groups that had offers from multiple Exchanges. This provides visibility into how often deduplication produces savings.
-
-## Content Attestation Levels (v1.0)
-
-In v1.0, content attestation level replaces the former content verification concept. Attestation level is both a deduplication tiebreaker and a selection ranking factor:
-
-| Level | Attestation | Guarantee |
-|---|---|---|
-| 0 | None | No attestation. Content may carry identifiers (DOI, IPTC GUID) for dedup, but nothing is cryptographically verifiable. |
-| 1 | Self-attested (provider) | Provider signed claims with Ed25519. Agent can recompute content hash from delivered bytes. |
-| 2 | Third-party (verification vendor) | Independent verification. Vendor crawled and measured the content. Agent trusts the attestation without re-verifying. |
-
-Higher attestation levels rank higher in the [Selection Engine](/components/broker/selection-engine/) because they provide stronger guarantees. When two offers for the same content have different attestation levels, the higher level is preferred even at slightly higher cost, because it enables automated dispute resolution and reduces the risk of receiving content that does not match what was advertised.
-
-Content identifiers (DOI, ISBN, IPTC GUID) used for deduplication are orthogonal to attestation level. A Level 0 article may carry a DOI -- that identifies WHAT content it is but does NOT verify that the delivered bytes match the identified content.
diff --git a/website/src/content/docs/components/broker/deployment.mdx b/website/src/content/docs/components/broker/deployment.mdx
deleted file mode 100644
index 51567e05..00000000
--- a/website/src/content/docs/components/broker/deployment.mdx
+++ /dev/null
@@ -1,339 +0,0 @@
----
-title: "Broker Deployment"
-description: "Deployment models, configuration reference, caching strategy, failure modes, and scaling considerations for the FORA Broker."
----
-
-import DeferredSdkNotice from '../../../../components/DeferredSdkNotice.mdx';
-
-
-
-## Deployment Models
-
-The Broker is designed for three deployment modes. The core logic (selection, budget) is identical across all three; what changes is the process model, configuration surface, and what the agent is responsible for.
-
-```mermaid
-graph LR
- subgraph "Model A: Agent SDK (embedded library)"
- AgentSDK[AI Agent + FORA SDK] --> MPAS[Exchanges]
- end
-
- subgraph "Model B: Sidecar Binary"
- AgentA[AI Agent] -->|localhost:9090| SidecarA[Broker Go binary]
- SidecarA --> MPA[Exchanges]
- end
-
- subgraph "Model C: Hosted SaaS"
- AgentC1[Agent 1] --> Hosted[Broker SaaS multi-tenant]
- AgentC2[Agent 2] --> Hosted
- Hosted --> MPC[Exchanges]
- end
-```
-
-### Model Comparison
-
-| Aspect | Agent SDK | Sidecar Binary | Hosted SaaS |
-|---|---|---|---|
-| **Process model** | In-process library | Separate process, same host | Remote service, separate infra |
-| **Language** | Go / TypeScript / Python | Go binary | Go service |
-| **Communication** | Function calls | Connect RPC over localhost | Connect RPC over HTTPS |
-| **Budget state** | In-process memory + local file | In sidecar memory | Server-side, per-tenant |
-| **Configuration** | Struct literal in code | YAML file + env vars | API |
-| **Multi-tenancy** | N/A (one license per instance) | N/A (one agent per sidecar) | Required -- session isolation |
-| **Latency overhead** | 0ms (in-process) | ~1ms (localhost RPC) | ~5-20ms (network hop) |
-| **Content fetch** | SDK fetches signed URL | Agent fetches signed URL | Agent fetches signed URL |
-| **Usage reporting** | The SDK files it in the background, signed with the agent's key | The agent files it with the issuing Exchange | The agent files it with the issuing Exchange |
-| **Who operates** | Agent developer | Agent operator | Broker SaaS provider |
-
-### Recommended Deployment
-
-**For agent developers (lowest friction)**: Ship the [Agent SDK](/components/agent-sdk/overview/). Agent developers add one dependency, pass their license key and budget, and call `Fetch`. No sidecar, no YAML, no infrastructure.
-
-**For agent operators (clear process boundary)**: Ship the sidecar binary. Operators run it alongside their agent, configure via YAML, and get multi-exchange comparison without code changes.
-
-**At scale (managed service)**: Offer hosted SaaS for teams that don't want to operate infrastructure. Multi-tenancy enables serving multiple AI companies from a single deployment.
-
-### Sidecar Binary
-
-```go
-func main() {
- cfg := loadConfig("broker.yaml")
- orch := broker.New(cfg)
-
- mux := http.NewServeMux()
- path, handler := forav1connect.NewBrokerServiceHandler(orch)
- mux.Handle(path, handler)
-
- http.ListenAndServe("localhost:9090", mux)
-}
-```
-
-### Hosted SaaS (Multi-Tenancy)
-
-The hosted Broker serves multiple AI companies. Tenant isolation is by API key or the verified agent identity (domain/key). Each tenant gets isolated: Exchange registry (different trust levels per tenant), budget tracking (per-tenant Redis keyspace), selection policy (per-tenant configuration), and selection logs.
-
-```go
-func main() {
- tenantStore := loadTenantStore()
-
- handler := &multiTenantBroker{
- tenants: tenantStore,
- }
-
- mux := http.NewServeMux()
- path, h := forav1connect.NewBrokerServiceHandler(handler)
- mux.Handle(path, authMiddleware(h))
-
- http.ListenAndServe(":443", mux)
-}
-
-func (m *multiTenantBroker) Resolve(
- ctx context.Context,
- req *connect.Request[forav1.DiscoveryRequest],
-) (*connect.Response[forav1.DiscoveryResponse], error) {
- tenantID := tenantFromContext(ctx)
- orch := m.tenants.GetBroker(tenantID)
- return orch.Resolve(ctx, req)
-}
-```
-
-### What Changes Between Modes
-
-The core interfaces (`ExchangeRegistry`, `SelectionEngine`, `BudgetTracker`) are identical. What changes is persistence strategy, configuration surface, and process lifecycle.
-
-## Configuration Reference
-
-```yaml
-# broker.yaml -- full configuration reference
-
-# Protocol version
-version: "1.0"
-
-# Exchange registry
-exchanges:
- - domain: exchange.ssp-alpha.com
- endpoint: https://exchange.ssp-alpha.com/fora/v1
- trust: verified
- priority: 1
- - domain: exchange.ssp-beta.com
- endpoint: https://exchange.ssp-beta.com/fora/v1
- trust: verified
- priority: 2
-
-# fora.json crawling (automated discovery)
-discovery:
- enabled: true
- domains:
- - techcrunch.com
- - nytimes.com
- - apnews.com
- interval: 1h
- # Trust level for auto-discovered Exchanges
- default_trust: discovered
-
-# Parallel fanout
-fanout:
- timeout: 60ms # Overall fanout deadline
- per_exchange_timeout: 50ms
- max_concurrency: 20 # Max concurrent DiscoverResources RPCs
-
-# Selection engine
-selection:
- policy: lowest_unit_cost # lowest_unit_cost | preferred_exchange | freshness_weighted
- verify_signatures: true # Verify exchange_signature on offers
-
-# Budget defaults (overridden per-session by DiscoveryRequest)
-budget:
- default_session_limit: 1.00
- default_currency: USD
-
-# Caching
-cache:
- fora_json_ttl: 1h
- exchange_health_ttl: 30s
- supply_response_ttl: 0s # Disabled by default
-
-# HTTP client
-http:
- tls_min_version: "1.3"
- idle_conn_timeout: 90s
- max_idle_conns_per_host: 10
-
-# Server (sidecar mode only)
-server:
- address: "localhost:9090"
- # No TLS for localhost sidecar. Add TLS config for network-exposed deployments.
-
-# Metrics
-metrics:
- enabled: true
- address: ":9091" # Prometheus metrics endpoint
-```
-
-## Exchange Registry
-
-The registry maintains a live list of known Exchanges, their endpoints, health status, and trust level.
-
-### Discovery Sources
-
-```mermaid
-graph LR
- subgraph "Discovery Sources"
- Manual[Manual Config YAML / env vars]
- ParseJSON[fora.json Crawling provider domains]
- OnDemand[Background Discovery on first request]
- Peer[Peer Exchange future]
- end
-
- Manual --> Registry
- ParseJSON --> Registry
- OnDemand --> Registry
- Peer -.-> Registry
-
- Registry[Exchange Registry unified view]
-```
-
-**Source 1: Manual configuration.** YAML file or environment variables listing known Exchange endpoints.
-
-**Source 2: fora.json crawling.** The Broker periodically crawls `/.well-known/fora.json` from provider domains to discover authorized Exchanges.
-
-**Source 3: Background discovery on first request.** When the agent requests a URI for an unknown domain, the Broker background-crawls that domain's fora.json. The current request is NOT blocked.
-
-**Source 4: Peer exchange (future).** Not designed for v1.
-
-### Trust Model
-
-| Trust Level | How Achieved | Allowed Operations |
-|---|---|---|
-| `discovered` | Found in a fora.json | DiscoverResources only (comparison shopping) |
-| `verified` | Operator has reviewed and approved | DiscoverResources + ExecuteTransaction |
-| `preferred` | Operator has marked as preferred | All operations + priority in selection |
-| `blocked` | Operator has explicitly blocked | No operations |
-
-An attacker who poisons a provider's fora.json would only gain `discovered` trust, which allows supply queries but not financial transactions. The operator must explicitly promote to `verified` before the Broker will spend money there.
-
-### Refresh Strategy
-
-| Data | Refresh Interval | Mechanism |
-|---|---|---|
-| Manual config | On startup + SIGHUP | File watch or signal handler |
-| fora.json per domain | 1 hour (configurable) | Background goroutine |
-| Exchange health | Every 30 seconds | Background check: a fresh fetch of the Exchange's `fora.json`, then a GET of the `health_endpoint` it names (200 means serving); with none named, the successful manifest fetch is the answer |
-| Trust levels | Manual only | Operator action via API or config |
-
-### Registry Interface
-
-```go
-type ExchangeRegistry interface {
- Resolve(constraints *forav1.RequestConstraints) []ExchangeEntry
- ResolveAll() []ExchangeEntry
- Register(entry ExchangeEntry)
- UpdateHealth(domain string, status HealthStatus)
- SetTrust(domain string, trust TrustLevel) error
-}
-```
-
-## Caching Strategy
-
-### Cache Layers
-
-```mermaid
-graph TB
- subgraph "Broker Caches"
- C1[fora.json Cache per provider domain TTL: 1 hour]
- C2[Exchange Health Cache per exchange TTL: 30 seconds]
- C3[ResourceResponse Cache per query fingerprint TTL: 5 seconds]
- end
-```
-
-| Cache | Key | TTL | Rationale |
-|---|---|---|---|
-| **fora.json** | Provider domain | 1 hour | fora.json changes rarely. Matches `max-age` from the provider's HTTP cache headers if present. |
-| **Exchange health** | Exchange domain | 30 seconds | Health can change rapidly. 30s balances freshness with probe cost. |
-| **ResourceResponse** | SHA-256 of (exchange + aisystem.name + uri list) | 5 seconds max | Only useful when multiple requests for the same content arrive in a burst. Set to 0 to disable. |
-| **Exchange public keys** | Exchange domain | 24 hours | Public keys for offer signature verification. Rotate slowly (90-day cycle). |
-
-Most caches are in-memory LRU. Redis is used only for shared state (per-period budget persistence) in multi-instance deployments.
-
-### ResourceResponse Caching
-
-ResourceResponse caching is disabled by default. Enable it only when:
-
-- The agent makes rapid successive requests for the same content (burst pattern)
-- The Broker is in hosted SaaS mode and multiple agents request the same content
-
-Even with caching enabled, 5 seconds is the maximum TTL. Offers can include `expires_at` timestamps; the cache respects these.
-
-## Failure Modes
-
-### Failure Taxonomy
-
-```mermaid
-graph TD
- F1[All Exchanges Timeout] -->|Result| R1[Return error to agent agent can retry or go direct]
- F2[Preferred Exchange Down] -->|Result| R2[Fallback to remaining Exchanges automatically]
- F3[Broker Crash Mid-Transaction] -->|Result| R3[Agent retries with same idempotency key, on any route]
- F4[Budget Tracking Lost on Crash] -->|Result| R4[Agent re-declares budget Exchange is authoritative]
- F5[Partial Exchange Failures] -->|Result| R5[Proceed with partial results, best-effort]
-```
-
-### Detailed Failure Handling
-
-**All Exchanges timeout.** The Broker returns a Connect error with code `UNAVAILABLE`. The agent can retry, query Exchanges directly, or fall back to non-FORA content access.
-
-**Preferred Exchange down.** The registry marks it as unhealthy. The selection engine picks from whatever responded. No special handling needed.
-
-**Broker crash mid-transaction.**
-
-| Crash Point | Impact | Recovery |
-|---|---|---|
-| Before ExecuteTransaction sent | No financial impact | Agent retries the full request |
-| After ExecuteTransaction sent, before response | Exchange has charged but agent doesn't know | Agent retries with the same `TransactionRequest.idempotency_key`, through this Broker once it is back, another Broker or directly -- each Exchange returns the same response (idempotency) |
-| After TransactionResponse returned | Agent has the signed URL | Agent tracks its reporting obligations and files its own reports |
-
-**Budget tracking lost on crash.** Session budget is ephemeral. The agent re-declares on reconnect. The Exchange's billing adapter is the authoritative source.
-
-**Partial Exchange failures.** This is the normal case. Metrics record which Exchanges failed.
-
-### Idempotency
-
-The Broker never generates `TransactionRequest.idempotency_key`. The key is the agent's, and the Broker forwards it unchanged on every sub-request: it cannot do otherwise, because each `AgentAcceptance` and the `AgentRequestAcceptance` sign it.
-
-Each Exchange scopes the key to the agent, whatever the route, so two agents behind one Broker that choose the same key do not collide. When the agent retries the same purchase with the same key, through this Broker, another one or directly, each Exchange recognizes a replay of its sub-request and answers from its stored result. When a key is reused with different items, or the same items in another order, the Exchange refuses the sub-request with `already_exists`, and the Broker reports that refusal on each affected item's `refusal`.
-
-## Scaling Considerations
-
-### Security Considerations
-
-| Threat | Attack | Countermeasure |
-|---|---|---|
-| Altering offers in transit | Broker raises a price or changes the terms of a relayed offer | The Exchange signs every offer field. A changed offer fails its signature, and the agent's acceptance signs the offer it chose. |
-| Steering discovery toward favored Exchanges | Broker has a side deal with an Exchange | Not stopped by the protocol. The agent can query Exchanges directly and buy there. |
-| Collecting agents' queries | Broker logs queries and sells data | Not stopped by the protocol. A data-handling agreement covers it, or the agent operator runs its own Broker. |
-
-### TLS Requirements
-
-All Exchange communication uses TLS 1.3 minimum. The Broker's HTTP clients are configured with `tls.Config{MinVersion: tls.VersionTLS13}`.
-
-### Exchange Public Keys
-
-The Broker fetches the Exchange's offer-signing keys from its WBA directory (the JWK Set at `/.well-known/http-message-signatures-directory`). It caches each Exchange's keys in memory, per Exchange domain, for five minutes and never past a key's `not_after`.
-
-Key rotation: an Exchange publishes the new key beside the old one for the grace period [Key Rotation](/protocol/authentication/#key-rotation) describes, so offers signed with either key verify.
-
-### Metrics
-
-The Broker exposes Prometheus-compatible metrics at `/metrics`:
-
-| Metric | Type | Description |
-|---|---|---|
-| `broker_exchange_latency_seconds` | Histogram | DiscoverResources response time per Exchange |
-| `broker_exchange_health` | Gauge | Health status (0=down, 1=degraded, 2=healthy) |
-| `broker_exchange_offer_count` | Histogram | Offers per DiscoverResources response |
-| `broker_selection_offers_total` | Counter | Offers at each pipeline stage |
-| `broker_dedup_sets` | Histogram | Content groups with multiple offers |
-| `broker_winning_unit_cost` | Histogram | unit_cost of the selected offer |
-| `broker_unit_cost_savings_ratio` | Histogram | Runner-up vs winning unit_cost |
-| `broker_budget_utilization_ratio` | Histogram | Session budget utilization |
-| `broker_budget_exhausted_total` | Counter | URIs answered with `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED` because the budget left no offer |
-| `broker_request_latency_seconds` | Histogram | Total DiscoveryRequest handling time |
-| `broker_requests_total` | Counter | Total DiscoveryRequests handled |
diff --git a/website/src/content/docs/components/broker/overview.mdx b/website/src/content/docs/components/broker/overview.mdx
index 86c3f785..1319e076 100644
--- a/website/src/content/docs/components/broker/overview.mdx
+++ b/website/src/content/docs/components/broker/overview.mdx
@@ -92,6 +92,16 @@ When at least one URL has an offer, every requested URL comes back as one `Offer
When no URL has an offer, or no publisher's `fora.json` can be read, the response carries no groups and states the reason once, in `DiscoveryResponse.absence_reason`. "No offer" is a successful answer, not an error: `BrokerService.Resolve` returns OK with the reason.
+## Order, budget and fees
+
+**Order.** A Broker returns the offers in its own order of relevance to the agent's request. Price is not a signal of that order: each offer carries its price and `Pricing.unit_cost`, and the agent compares them under its own policy. An agent names the Exchanges it has a relationship with, such as a subscription, in `RequestConstraints.preferred_exchanges`, and a Broker SHOULD put their offers first when they have the resource.
+
+**Same content, several Exchanges.** When several Exchanges sell the same content, the agent gets each offer as a full package, and `Offer.identity` tells it that they are the same content. A Broker groups such offers and drops none of them, so the agent compares the packages and decides which to buy.
+
+**Budget.** The agent's budget is its own constraint. The limits it states in `DiscoveryRequest.constraints`, such as `max_price` or `period_budget` with `budget_period`, narrow which offers come back, and a Broker applies them as a filter, not as an order. A URL the budget leaves with no offer is answered with `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED`. The charge is the Exchange's, at purchase.
+
+**Fees.** An offer carries only the Exchange's price, which the Exchange signed. A Broker's fee has no field: it is disclosed out of band and added to that price, never folded into it.
+
## Buying through the Broker
An agent commits to offers in a separate call, one of two ways. The execute target of each offer is `Offer.exchange`, which the Exchange signed, so a Broker cannot redirect a purchase.
@@ -135,6 +145,8 @@ The Broker forwards `Requester.scopes` unchanged and in order, because the Excha
## What a Broker never does
- **Charge.** `Resolve` never charges, and on a purchase each Exchange charges the agent's account.
+- **Choose for the agent.** It returns the offers in its order of relevance, and the agent decides which to buy.
+- **Buy or resell.** It never buys or resells under an account of its own.
- **Hold the agent's keys.** The agent, or the [Identity Service](/components/identity/overview/) for an agent on MCP, signs the acceptances.
- **Issue signed URLs.** The Exchange issues each delivery URL, and the Broker passes it on.
- **See the content.** The Broker is not on the delivery path: the agent fetches the resource from the provider's CDN.
@@ -144,9 +156,16 @@ The Broker forwards `Requester.scopes` unchanged and in order, because the Excha
Anyone may run a Broker, with or without an Exchange of their own. It is a role of its own, with its own key and its own `fora.json`, and each Exchange decides which Brokers it admits. A search company, for example, can run only a Broker, behind its own search.
-FORA Server's Broker is a separate service. It needs PostgreSQL, Redis, a public hostname with TLS, and a registry that lists the Exchanges its operator trusts. The address of each Exchange always comes from that Exchange's own `fora.json`, never from the registry. The [deployment guide](https://github.com/FORA-Protocol/reference-implementation/blob/main/src/broker/DEPLOYMENT.md) is in the FORA Server repository.
+FORA Server's Broker is a separate service, published as the container image `ghcr.io/fora-protocol/broker`. Its operator deploys it with:
+
+- PostgreSQL, Redis, and a public hostname with TLS;
+- two Ed25519 keys: an identity key, whose public half the Broker publishes in its key directory, and a relay key, which signs the Broker's requests to Exchanges;
+- a registry file that lists the Exchanges it routes to, each with a `trust_level`: `DISCOVERED` Exchanges answer discovery and are never paid, `VERIFIED` and `PREFERRED` ones also take purchases, and `BLOCKED` ones take neither;
+- a withdrawn-keys file, which the Broker publishes. A FORA Server Exchange is configured with one Broker (`EXCHANGE_BROKER_WELLKNOWN_URL`) and rejects signed requests while it can't read that Broker's list, so the Broker goes up first.
+
+The address of each Exchange always comes from that Exchange's own `fora.json`, never from the registry. The [deployment guide](https://github.com/FORA-Protocol/reference-implementation/blob/main/src/broker/DEPLOYMENT.md) is in the FORA Server repository.
-As of November 2026, Gutefrage runs the whole FORA stack itself, on its own servers, and has launched it on [helpster.de](https://www.helpster.de/), one of its properties. Its Broker searches the Elasticsearch index Gutefrage already runs, with a search template in which Gutefrage decides which fields are searched, how results rank and which filters agents may use.
+Gutefrage runs the whole FORA stack itself, on its own servers, and has launched it on [helpster.de](https://www.helpster.de/), one of its properties. Its Broker searches the Elasticsearch index Gutefrage already runs, with a search template in which Gutefrage decides which fields are searched, how results rank and which filters agents may use.
## Related pages
diff --git a/website/src/content/docs/components/broker/selection-engine.mdx b/website/src/content/docs/components/broker/selection-engine.mdx
deleted file mode 100644
index 95d4b73d..00000000
--- a/website/src/content/docs/components/broker/selection-engine.mdx
+++ /dev/null
@@ -1,554 +0,0 @@
----
-title: "Selection Engine"
-description: "How the Broker compares offers across Exchanges, deduplicates by content identity, ranks by cost, and selects the best deal."
----
-
-## Purpose
-
-The Selection Engine compares offers across Exchanges, deduplicates by content identity, ranks by cost, and applies constraints. This is the "auction logic" of the Broker -- it determines which Exchange wins a given content request.
-
-### Discovery-Mode Agnostic
-
-The Selection Engine operates on offers, not on how those offers were discovered. Whether the agent sent specific URIs (`DiscoveryRequest.uris`), structured search filters (`DiscoveryRequest.search_filters`), or a natural-language query (`DiscoveryRequest.query`), the Broker resolves all three into a set of offers from Exchanges. The Selection Engine then applies the same pipeline -- filter, deduplicate, rank, verify -- to every offer identically.
-
-This means the selection logic does not need to know or care about discovery mode. An offer discovered via structured search competes on equal footing with an offer from a specific URI request. The `OfferGroup.discovery_method` field records how each offer was found (for transparency and logging), but it does not influence ranking.
-
-## Multi-Exchange Query Fanout
-
-The core value proposition of the Broker is parallel supply discovery. The Broker fans out `DiscoverResources` to N Exchanges concurrently, collecting offers for comparison.
-
-### Concurrency Model
-
-Uses `errgroup.Group` with a semaphore for bounded concurrency and `context.WithTimeout` for per-Exchange deadlines. When the `DiscoveryRequest` contains multiple URIs, the Broker sends all URIs in a single `ResourceQuery` to each Exchange (not one query per URI). The Exchange returns `OfferGroups` -- one per URI -- so the Broker knows which offers map to which content.
-
-```go
-func (d *Discovery) FanOut(
- ctx context.Context,
- exchanges []ExchangeEntry,
- req *forav1.DiscoveryRequest,
-) []FanOutResult {
-
- // Hard deadline: don't wait longer than this for any Exchange.
- ctx, cancel := context.WithTimeout(ctx, d.fanoutTimeout) // default: 60ms (p99 target)
- defer cancel()
-
- results := make([]FanOutResult, len(exchanges))
- g, gctx := errgroup.WithContext(ctx)
- g.SetLimit(d.maxConcurrency) // default: 20
-
- for i, mp := range exchanges {
- i, mp := i, mp
- g.Go(func() error {
- query := toResourceQuery(req, mp) // includes all req.Uris
- start := time.Now()
-
- resp, err := d.clients[mp.Domain].DiscoverResources(gctx, connect.NewRequest(query))
-
- results[i] = FanOutResult{
- Exchange: mp,
- Response: resp,
- Err: err,
- Latency: time.Since(start),
- }
-
- // Don't return error -- we want partial results, not all-or-nothing.
- return nil
- })
- }
-
- g.Wait()
- return results
-}
-```
-
-### Time Budget in ResourceQuery
-
-`ResourceQuery` carries an optional `deadline` field (default: 500ms). This informs the Exchange how much time the agent has allocated for the entire discovery phase. Exchanges SHOULD prioritize speed when the deadline is tight.
-
-```go
-func (d *Discovery) perExchangeDeadline(mp ExchangeEntry, requestDeadline time.Duration) time.Duration {
- p99 := d.latencyTracker.P99(mp.Domain)
- if p99 == 0 {
- return requestDeadline // no history, use full budget
- }
- // Allow 2x historical p99, capped at request deadline
- budget := min(p99*2, requestDeadline)
- return budget
-}
-```
-
-### Timeout Handling
-
-Three timeout layers, each independently enforced:
-
-```mermaid
-graph TD
- subgraph "Timeout Layers"
- T1["Layer 1: Per-Exchange context.WithTimeout per goroutine default: 50ms"]
- T2["Layer 2: Fanout deadline context.WithTimeout on errgroup default: 60ms"]
- T3["Layer 3: Request deadline agent's context deadline forwarded from DiscoveryRequest"]
- end
-
- T3 -->|cancels| T2
- T2 -->|cancels| T1
-```
-
-| Layer | Default | Purpose |
-|---|---|---|
-| Per-Exchange | 50ms | Individual slow Exchange doesn't block others |
-| Fanout deadline | 60ms | Overall discovery phase has a hard cap |
-| Request deadline | From agent's context | Agent controls the total request timeout |
-
-**Partial results are normal.** If 3 of 5 Exchanges respond within the deadline, the Broker proceeds with 3 responses. Timed-out Exchanges are recorded in metrics and their health status is degraded.
-
-### Client Pool
-
-Pre-established HTTP/2 connections to each Exchange. Connect-Go clients are created at startup and reused.
-
-```go
-type ClientPool struct {
- mu sync.RWMutex
- clients map[string]forav1connect.ExchangeServiceClient
- signer helpers.Signer // the Broker's own request-signing key
-}
-
-func (p *ClientPool) GetOrCreate(mp ExchangeEntry) forav1connect.ExchangeServiceClient {
- p.mu.RLock()
- if c, ok := p.clients[mp.Domain]; ok {
- p.mu.RUnlock()
- return c
- }
- p.mu.RUnlock()
-
- p.mu.Lock()
- defer p.mu.Unlock()
-
- // Double-check after acquiring write lock.
- if c, ok := p.clients[mp.Domain]; ok {
- return c
- }
-
- base := &http.Transport{
- MaxIdleConnsPerHost: 10,
- IdleConnTimeout: 90 * time.Second,
- ForceAttemptHTTP2: true,
- TLSClientConfig: &tls.Config{MinVersion: tls.VersionTLS13},
- }
- httpClient := &http.Client{
- // The Broker signs every request it sends alone, naming its own key
- // directory in its Signature-Agent member.
- Transport: core.NewSigningTransport(p.signer, base,
- core.WithSignatureAgent("https://broker.example")),
- Timeout: 5 * time.Second, // connection-level timeout
- }
-
- client := forav1connect.NewExchangeServiceClient(httpClient, mp.Endpoint)
- p.clients[mp.Domain] = client
- return client
-}
-```
-
-## Selection Pipeline
-
-### Single-URI Pipeline
-
-For single-URI queries, five stages are executed sequentially on the combined offer set:
-
-```mermaid
-graph LR
- A[Collect all Offers from N Exchanges] --> B[One offer per term agent chooses an honourable offer]
- B --> C[Filter by RequestConstraints]
- C --> D[Deduplicate by ResourceIdentity]
- D --> E[Rank by unit_cost lowest wins]
- E --> F[Verify exchange_signature]
- F --> G[Winner]
-```
-
-### Multi-URI Batch Pipeline
-
-For multi-URI batch queries, selection is 2-dimensional: the engine selects the best offer per URI, then checks whether the total cost across all URIs fits within budget.
-
-```mermaid
-graph LR
- A[Collect OfferGroups from N Exchanges] --> B[For each URI: merge offers across MPs]
- B --> C[Per-URI: filter, dedup, rank]
- C --> D[Select best offer per URI]
- D --> E[Sum costs across all URIs]
- E --> F{Total within budget?}
- F -->|Yes| G[Return one OfferGroup per URI]
- F -->|No| H[Drop lowest-priority URIs until budget fits]
-```
-
-The batch selection algorithm:
-
-1. For each URI, collect offers from all Exchange responses (flatten OfferGroups with matching URI).
-2. Within each URI group, run the full single-URI pipeline (filter, dedup, rank).
-3. Select the best offer per URI.
-4. Sum costs across all selected offers to get the batch total.
-5. If the total exceeds budget, drop offers for the lowest-priority URIs (by position in the original request) until the total fits. A URI left with no offer is answered with `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED` on its `OfferGroup`.
-6. Return one `OfferGroup` per URI. `Resolve` is discovery only: it charges nothing. The agent buys the selected offers in a separate call.
-
-### Selection Interface
-
-```go
-type SelectionEngine interface {
- // Select handles single-URI queries. Returns one winning offer.
- Select(
- ctx context.Context,
- responses []FanOutResult,
- constraints *forav1.RequestConstraints,
- budget BudgetTracker,
- ) (*Selection, error)
-
- // SelectBatch handles multi-URI queries. Returns one winning offer per URI,
- // with total cost checked against budget.
- SelectBatch(
- ctx context.Context,
- responses []FanOutResult,
- constraints *forav1.RequestConstraints,
- budget BudgetTracker,
- ) (*BatchSelection, error)
-}
-
-type Selection struct {
- Exchange ExchangeEntry
- Offer *forav1.Offer
- Rationale SelectionRationale
-}
-
-type BatchSelection struct {
- Selections []Selection
- TotalCost float64
- Currency string
- Rationale BatchSelectionRationale
-}
-
-type SelectionRationale struct {
- TotalOffers int
- WithHonourableTerm int // offers exposing a term whose restrictions the agent can honour
- AfterConstraints int
- AfterDedup int
- WinningUnitCost float64
- RunnerUpUnitCost float64 // 0 if no runner-up
- DeduplicatedSets int // how many content groups had multiple offers
- RejectedReasons map[string]int // reason -> count
-}
-```
-
-## Stage 1: Surface Offer Restrictions (Agent Chooses an Offer)
-
-Restrictions ride on the offer, not on a broker-side filter. Each offer's single `terms[0]` entry is a `LicenseTerm` carrying `repeated Restriction` (with kinds like `RESTRICTION_KIND_FUNCTION`, `RESTRICTION_KIND_GEOGRAPHY`, `RESTRICTION_KIND_USER_TYPE`). The Broker does **not** match a requester's self-declared attributes (function, geography, user type) against these restrictions to drop offers. Instead it surfaces the restrictions carried on each offer's own term; the agent chooses the offer whose restrictions it can honour. Real enforcement happens downstream at accept -> report -> reconcile.
-
-The Broker may help the agent by annotating which term it can honour, but it does not silently remove offers whose terms carry restrictions. The function-token vocabulary is the hyphenated string form (`ai-train`, `ai-index`, `ai-input`, `crawl`, `search`, ...).
-
-```go
-// honourableTerm returns the first LicenseTerm whose restrictions the agent
-// can honour, given its declared function. It does NOT drop the offer when no
-// term matches -- the offer is still surfaced; the agent decides.
-func honourableTerm(offer *forav1.Offer, declaredFunction string) *forav1.LicenseTerm {
- for _, term := range offer.Terms {
- if restrictionsHonoured(term.Restrictions, declaredFunction) {
- return term
- }
- }
- return nil
-}
-
-func restrictionsHonoured(restrictions []*forav1.Restriction, declaredFunction string) bool {
- for _, r := range restrictions {
- if r.Kind != forav1.RestrictionKind_RESTRICTION_KIND_FUNCTION {
- continue // geography / user-type evaluated by the agent against its own context
- }
- // If prohibited contains the declared function, this term cannot be honoured.
- for _, prohibited := range r.Prohibited {
- if prohibited == declaredFunction {
- return false
- }
- }
- // If permitted is non-empty, the declared function must be in it.
- if len(r.Permitted) > 0 {
- found := false
- for _, perm := range r.Permitted {
- if perm == declaredFunction {
- found = true
- break
- }
- }
- if !found {
- return false
- }
- }
- }
- return true
-}
-```
-
-## Stage 2: Filter by RequestConstraints
-
-The agent's `DiscoveryRequest.constraints` specifies budget and preference limits.
-
-```go
-func filterByConstraints(offers []rankedOffer, c *forav1.RequestConstraints, budget BudgetTracker) []rankedOffer {
- var passed []rankedOffer
- for _, o := range offers {
- p := o.Offer.Pricing
-
- // The price this Broker counts against the budget. Which price that is
- // is the Broker's own choice; the protocol does not specify it.
- price := budgetPrice(p)
-
- // max_price: drop offers above it
- if c.MaxPrice != nil && price > c.MaxPrice.Amount {
- continue
- }
-
- // max_unit_cost: reject offers with unit_cost above threshold
- if c.MaxUnitCost != nil && p.UnitCost != nil && *p.UnitCost > *c.MaxUnitCost {
- continue
- }
-
- // reporting_capable: if agent can't report, skip offers that require it
- if c.ReportingCapable != nil && !*c.ReportingCapable {
- if o.Offer.Reporting != nil && o.Offer.Reporting.Required {
- continue
- }
- }
-
- // delivery_preference: filter by supported delivery methods
- if len(c.DeliveryPreference) > 0 {
- if !deliveryMethodAccepted(o.Offer.DeliveryMethod, c.DeliveryPreference) {
- continue
- }
- }
-
- // Period and session budget: drop the offer if buying it would take the
- // spend past the budget. This filter reserves nothing.
- if !budget.CanAfford(price, p.Currency) {
- continue
- }
-
- passed = append(passed, o)
- }
- return passed
-}
-```
-
-When the budget drops every offer for a URI, that URI is answered with `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED`, a successful "no result". A budget is never reported as `OFFER_ABSENCE_REASON_NOT_AUTHORIZED`, and never as an error such as `RESOURCE_EXHAUSTED`.
-
-## Stage 3: Deduplicate by ResourceIdentity
-
-The same article may be offered by multiple Exchanges (e.g., an AP wire story available from Exchange A and Exchange B). The Broker groups offers by resource identity and keeps only the cheapest per group.
-
-```go
-// contentIdentityKey produces a deduplication key from layered identity fields.
-// Matching priority: iptc_guid > doi > content_hash > canonical_url
-func contentIdentityKey(id *forav1.ResourceIdentity) string {
- if id == nil {
- return uuid.NewString() // no identity = unique (no dedup)
- }
- if id.IptcGuid != nil && *id.IptcGuid != "" {
- return "iptc:" + *id.IptcGuid
- }
- if id.Doi != nil && *id.Doi != "" {
- return "doi:" + *id.Doi
- }
- if id.ContentHash != nil && *id.ContentHash != "" {
- return "hash:" + *id.ContentHash
- }
- if id.CanonicalUrl != nil && *id.CanonicalUrl != "" {
- return "url:" + *id.CanonicalUrl
- }
- return uuid.NewString()
-}
-```
-
-For the full deduplication logic, see the [Content Deduplication](/components/broker/content-dedup/) page.
-
-## Stage 4: Multi-Factor Ranking
-
-After deduplication, remaining offers are ranked by a multi-factor scoring engine. "Lowest unit_cost wins" is the default policy but not the only option. The selection engine considers:
-
-1. **Subscription content (zero marginal cost).** Offers with `subscription_id` set rank above all per-request offers. The selection engine discovers subscriptions when Exchange responses include offers with `subscription_id` populated.
-
-2. **Agent's preferred Exchanges.** The agent declares `preferred_exchanges` in `RequestConstraints`. Ranking priority: (a) subscription offers from preferred Exchanges, (b) subscription offers from any Exchange, (c) per-request offers from preferred Exchanges, (d) per-request offers ranked by unit_cost.
-
-3. **Provider reputation scores.** Exchanges that consistently deliver accurate, well-structured content rank higher.
-
-4. **Content attestation level (v1.0).** Level 2 (third-party attested) > Level 1 (self-attested) > Level 0 (no attestation). Offers with `ResourceAttestation` entries from trusted verification vendors rank higher because they provide independently verifiable claims about the content.
-
-5. **Quantity estimation accuracy.** Exchanges with historically accurate estimated_quantity values rank higher.
-
-6. **Business relationships.** The agent declares known relationships via `RequestConstraints`.
-
-```go
-func (e *defaultEngine) rankOffers(offers []rankedOffer, constraints *forav1.RequestConstraints) []rankedOffer {
- for i := range offers {
- score := 0.0
-
- // Zero marginal cost: subscription offers always win.
- if offers[i].Offer.SubscriptionId != nil && *offers[i].Offer.SubscriptionId != "" {
- score += 10000.0 // effectively infinite preference
- // Extra boost for subscription offers from preferred exchanges
- if isPreferred(offers[i].Exchange.Domain, constraints.PreferredExchanges) {
- score += 5000.0
- }
- }
-
- // unit_cost: lower is better (inverted for scoring)
- uc := unitCostOf(offers[i])
- if uc > 0 && uc < math.MaxFloat64 {
- score += 1000.0 / uc
- }
-
- // Preferred exchange bonus
- if isPreferred(offers[i].Exchange.Domain, constraints.PreferredExchanges) {
- score += 500.0
- }
-
- // Content attestation level bonus (v1.0)
- // Level 2 (third-party) > Level 1 (self-attested) > Level 0 (none)
- score += float64(attestationLevel(offers[i].Offer)) * 100.0
-
- // Quantity estimation accuracy bonus (historical)
- score += e.quantityAccuracy.Score(offers[i].Exchange.Domain) * 50.0
-
- offers[i].Score = score
- }
-
- sort.Slice(offers, func(i, j int) bool {
- return offers[i].Score > offers[j].Score // descending: highest score wins
- })
- return offers
-}
-
-func unitCostOf(o rankedOffer) float64 {
- if o.Offer.Pricing.UnitCost != nil {
- return *o.Offer.Pricing.UnitCost
- }
- // Fallback: compute from rate and estimated_quantity
- if o.Offer.Pricing.EstimatedQuantity != nil && *o.Offer.Pricing.EstimatedQuantity > 0 {
- return o.Offer.Pricing.Rate / float64(*o.Offer.Pricing.EstimatedQuantity)
- }
- return math.MaxFloat64 // no unit_cost available, rank last
-}
-```
-
-## Stage 5: Verify exchange_signature
-
-The Broker SHOULD verify `exchange_signature` on the winning Offer before proceeding. This prevents an Exchange from tampering with its own offer fields after signing.
-
-```go
-func verifyOfferSignature(offer *forav1.Offer, exchange string, pubKeys map[string]ed25519.PublicKey) error {
- if offer.Signature == "" {
- return fmt.Errorf("offer signature is required but empty")
- }
-
- key, ok := pubKeys[exchange]
- if !ok {
- return fmt.Errorf("no public key for exchange %q", exchange)
- }
-
- return helpers.VerifyOffer(offer, offer.Signature, key)
-}
-```
-
-## Pluggable Selection Policies
-
-The default selection policy is "lowest unit_cost wins." Production deployments may want custom policies:
-
-```go
-type SelectionPolicy interface {
- // Rank orders offers by preference. Index 0 is the winner.
- Rank(offers []rankedOffer, constraints *forav1.RequestConstraints) []rankedOffer
-}
-
-// Built-in policies
-type LowestUnitCostPolicy struct{} // Default: cheapest wins
-type PreferredExchangePolicy struct{} // Prefer specific Exchanges, then unit_cost
-type FreshnessWeightedPolicy struct{} // Prefer newer content, weighted against unit_cost
-type QualityScorePolicy struct{} // Use Exchange reputation scores
-```
-
-The policy is set at Broker construction time, not per-request. Per-request behavior is controlled through `RequestConstraints`.
-
-## Exchange Authorization Verification
-
-Before executing a transaction, the Broker SHOULD verify that the Exchange is authorized by the provider whose content is being licensed. This is the FORA equivalent of ads.txt verification in programmatic advertising.
-
-```go
-func (o *Broker) verifyExchangeAuthorization(
- ctx context.Context,
- contentDomain string,
- exchangeDomain string,
-) (bool, error) {
- manifest, err := o.parseJSONCache.Get(ctx, contentDomain)
- if err != nil {
- return false, fmt.Errorf("fetching fora.json for %s: %w", contentDomain, err)
- }
- for _, authorized := range manifest.AuthorizedExchanges {
- if authorized.Domain == exchangeDomain {
- return true, nil
- }
- }
- return false, nil // Exchange not authorized by provider
-}
-```
-
-This check runs after offer selection, before `ExecuteTransaction`. If verification fails, the Broker falls back to the next-ranked offer.
-
-## Selection Rationale Logging
-
-Every selection decision is logged with full rationale for audit:
-
-```go
-type SelectionLog struct {
- // Broker-internal audit key correlating this decision to the inbound request.
- // Request/response correlation rides the X-Request-ID HTTP header, not a message field.
- FORARequestID string `json:"fora_request_id"`
- Timestamp time.Time `json:"timestamp"`
- ExchangesQueried []string `json:"exchanges_queried"`
- ExchangesResponded []string `json:"exchanges_responded"`
- TotalOffers int `json:"total_offers"`
- FilteredOffers int `json:"filtered_offers"`
- DeduplicatedSets int `json:"deduplicated_sets"`
- WinningExchange string `json:"winning_exchange"`
- WinningOfferID string `json:"winning_offer_id"`
- WinningUnitCost float64 `json:"winning_unit_cost"`
- RunnerUpUnitCost float64 `json:"runner_up_unit_cost,omitempty"`
- RejectionReasons map[string]int `json:"rejection_reasons"`
- SelectionPolicy string `json:"selection_policy"`
- LatencyMs int64 `json:"latency_ms"`
-}
-```
-
-## Transparency and Auditability
-
-The Broker is an intermediary. FORA addresses transparency through:
-
-### Signed Offer Pass-Through
-
-The Exchange signs the Offer using its private key. The Broker passes the signed offer through to the agent. **The Broker cannot inflate prices** because the Exchange's signature covers the pricing fields.
-
-```mermaid
-sequenceDiagram
- participant MP as Exchange
- participant Orch as Broker
- participant Agent as AI Agent
-
- MP->>Orch: ResourceResponse (Offer with exchange_signature)
- Note over Orch: Broker CANNOT modify offer fields without invalidating signature
- Orch->>Agent: DiscoveryResponse (includes original signed Offer)
- Note over Agent: Agent verifies exchange_signature against MP's public key. Price tampering detected.
-```
-
-### Fee Disclosure
-
-`DiscoveryResponse` carries the signed Offers unchanged (one `OfferGroup` per requested URI); it has no `broker_fee` field, so the Broker cannot fold a fee into the discovery result. Any Broker fee is additive and disclosed out of band, never hidden in the offer price — the Exchange's signature over the Offer's pricing fields makes silent inflation detectable. The agent sees: Exchange price + Broker fee = total cost.
-
-### Audit Trail
-
-For full auditability, the agent operator should:
-
-1. **Verify offer signatures** on every DiscoveryResponse
-2. **Retain selection logs** to prove the Broker selected the cheapest offer
-3. **Spot-check by going direct** -- query Exchanges directly to verify the Broker isn't hiding better offers
diff --git a/website/src/content/docs/components/content-ingestion/sources.mdx b/website/src/content/docs/components/content-ingestion/sources.mdx
index 637fbf4a..bd5d8aec 100644
--- a/website/src/content/docs/components/content-ingestion/sources.mdx
+++ b/website/src/content/docs/components/content-ingestion/sources.mdx
@@ -325,7 +325,7 @@ Companies like GumGum, Peer39, and DoubleVerify already index the web for contex
### Provenance Tracking
Every catalog entry records who provided it and when, via two fields on `ResourceEntry`:
-- `provenance_source` -- e.g., "gumgum.com", "ssp-alpha.com", "wordpress-plugin"
+- `provenance_source` -- e.g., "gumgum.com", "alpha.example", "wordpress-plugin"
- `provenance_timestamp` -- when the metadata was collected/generated
This creates an audit trail for "where did this catalog entry come from?" -- essential when multiple parties can push metadata for the same content.
diff --git a/website/src/content/docs/components/edge-function/bot-detection.mdx b/website/src/content/docs/components/edge-function/bot-detection.mdx
index ff8311ba..29f72060 100644
--- a/website/src/content/docs/components/edge-function/bot-detection.mdx
+++ b/website/src/content/docs/components/edge-function/bot-detection.mdx
@@ -15,15 +15,15 @@ Three layers of bot detection, from cheapest to most expensive:
### Layer 1: User-Agent String Matching (< 0.01ms)
-Known AI crawler User-Agents. This is the primary detection mechanism. The AI companies run three kinds of bot, and the edge function treats each kind differently on a licensed path:
+Known AI crawler User-Agents. This is the primary detection mechanism. The AI companies run three kinds of bot. On a licensed path, the edge function refuses all three by default, and the provider can let an AI search crawler through:
| Kind | What it does | On a licensed path |
|---|---|---|
| Training and datasets | Collects pages to train models or build datasets | `403` |
| Live fetch | Loads a page because a person asked a question | `403` |
-| AI search | Builds the index that the vendor's AI search and answers draw on | The provider decides per crawler |
+| AI search | Builds the index that the vendor's AI search and answers draw on | `403` by default; the provider can let one through |
-A live fetch and a training crawl are what an agent buys a license for, so both get the `403` that points to the Exchange. An AI search crawler decides whether the site appears in that vendor's AI search at all. The vendors document that blocking one removes the site from those results, so the provider chooses which to let through. The default deny list refuses most of them, including OAI-SearchBot and PerplexityBot, and its generic `bot` token catches others; a provider that wants to stay in a vendor's AI search lets that crawler through with its own lists.
+A live fetch and a training crawl are what an agent buys a license for, so both get the `403` that points to the Exchange. An AI search crawler gets the same `403` by default. The vendors document that refusing one takes the site out of that vendor's AI search, so the provider decides: to keep its licensed paths in one vendor's AI search, it lets that vendor's crawler through (see [Changing the lists](#changing-the-lists)). Paths outside the licensed paths are not affected.
{[
'# Training and datasets: 403 on licensed paths',
@@ -33,24 +33,51 @@ A live fetch and a training crawl are what an agent buys a license for, so both
botLines('live-fetch'),
`${'fora-ai-buyer'.padEnd(24)}- a FORA agent that names itself`,
'',
- '# AI search: the provider decides',
+ '# AI search: 403 on licensed paths unless the provider lets it through',
botLines('ai-search'),
].join('\n')}
-The lists come from one file, `website/src/data/ai-bots.ts`, with each bot's vendor documentation; they are also published as [`/registry/ai-bots.json`](/registry/ai-bots.json) and shown, with what refusing each one costs, on [What you can do about AI crawlers today](/resources/ai-crawlers/).
+These names come from one file, `website/src/data/ai-bots.ts`, with each bot's vendor documentation; they are also published as [`/registry/ai-bots.json`](/registry/ai-bots.json) and shown, with what refusing each one costs, on [What you can do about AI crawlers today](/resources/ai-crawlers/).
-The edge function has built-in lists: a deny list of AI bots, which also matches the generic tokens `bot`, `crawler` and `spider`, and an allow list of search crawlers that read for free, such as Googlebot and Bingbot. An empty User-Agent is treated as a bot. A provider replaces either list with `BOT_UA_DENY_JSON` or `BOT_UA_ALLOW_JSON` (see [Edge function deployment](/components/edge-function/deployment/)). A bot moves between lists when its vendor changes what it does, and leaves them when its vendor retires it.
+#### The built-in lists
+
+FORA Server's edge function matches the User-Agent against two built-in lists of regular expressions, case-insensitively, on its first 512 characters. It checks the search-crawler allow list first: a caller that matches it passes. A caller that matches the AI-bot deny list gets the `403`. A caller that matches neither passes as a person, and a caller that sends no User-Agent counts as a bot.
+
+```
+# Search-crawler allow list: passes
+Googlebot bingbot DuckDuckBot Applebot(?!-Extended) YandexBot
+
+# AI-bot deny list: 403 on licensed paths
+bot\b crawler spider GPTBot ClaudeBot anthropic-ai
+OAI-SearchBot CCBot PerplexityBot Bytespider Google-Extended
+Meta-ExternalAgent Claude-User ChatGPT-User Perplexity-User Meta-ExternalFetcher
+```
+
+The deny list names `PerplexityBot` and `OAI-SearchBot`. Its generic patterns `bot\b`, `crawler` and `spider` catch most other bots that name themselves, AI search crawlers such as `Claude-SearchBot` among them. A bot whose User-Agent matches none of the patterns passes, so a provider that wants it refused adds it to the deny list.
+
+On Cloudflare, where the request carries Cloudflare's verified-bot category, the edge function reads it before the lists. A verified search engine crawler passes. A bot that Cloudflare verifies as AI Search, AI Crawler or AI Assistant gets the `403`, whatever the lists say.
+
+#### Changing the lists
+
+The provider changes the lists with two settings of the edge function (see [Edge function deployment](/components/edge-function/deployment/#what-the-edge-function-needs-to-know)):
+
+- `BOT_UA_ALLOW_JSON` replaces the search-crawler allow list. To let an AI search crawler through, the provider sets the built-in allow list plus that crawler's name, for example `["Googlebot","bingbot","DuckDuckBot","Applebot(?!-Extended)","YandexBot","OAI-SearchBot"]`. The allow list is checked first, so the name passes even though the deny list also matches it. The Cloudflare category still applies: a crawler that Cloudflare verifies as AI Search gets the `403` there.
+- `BOT_UA_DENY_JSON` replaces the AI-bot deny list, for example to add a bot that the built-in list misses.
+
+Each setting replaces its built-in list entirely and does not extend it, so a list the provider sets repeats the built-in entries it still wants. Each is a JSON array of 1 to 64 regular expressions, each at most 256 characters long. An empty list, a pattern that does not compile or a value that does not parse makes the edge function answer `500` to every request, so check the bot gate after each change. [`CONFIGURATION.md`](https://github.com/FORA-Protocol/reference-implementation/blob/v1.1.0/src/edge/CONFIGURATION.md) has the full rules.
+
+A bot moves between lists when its vendor changes what it does, and leaves them when its vendor retires it.
#### Search crawlers that also serve AI
-Googlebot, Bingbot and Applebot feed ordinary search and the AI features built on it: Google's AI Overviews and AI Mode, Bing's Copilot, Apple's Siri and Spotlight. The edge function never matches them, because refusing them takes the site out of search. These vendors treat their AI uses as allowed unless the site refuses them, so a provider refuses each use it has not offered with that vendor's own control, outside the edge function:
+Googlebot, Bingbot and Applebot feed ordinary search and the AI features built on it: Google's AI Overviews and AI Mode, Bing's Copilot, Apple's Siri and Spotlight. The edge function never refuses them: they are on its search-crawler allow list, because refusing them takes the site out of search. These vendors treat their AI uses as allowed unless the site refuses them, so a provider refuses each use it has not offered with that vendor's own control, outside the edge function:
- `Google-Extended` in `robots.txt`: Gemini training and grounding in Gemini Apps and Vertex AI. It does not affect Google Search.
- The "Search generative AI" setting in Google Search Console: AI Overviews, AI Mode and Discover.
- `Applebot-Extended` in `robots.txt`: Apple's model training. The `nosnippet` robots meta tag keeps a page out of Apple's AI answers.
- Bing's page-level `ai-training` and `ai-grounding` settings: Microsoft's training and Copilot grounding.
-`Google-Extended` and `Applebot-Extended` are `robots.txt` tokens only. No crawler sends them as a User-Agent, so the edge function does not list them. The same holds for the `__usage_ai-training` token of [draft-kelley-aipref-training-preference-signal](https://datatracker.ietf.org/doc/draft-kelley-aipref-training-preference-signal/), which refuses training to every crawler that honors it.
+`Google-Extended` and `Applebot-Extended` are `robots.txt` tokens only. No crawler sends them as a User-Agent, so a User-Agent list cannot refuse those uses, and the `Google-Extended` entry on the edge function's deny list matches no real crawler. The same holds for the `__usage_ai-training` token of [draft-kelley-aipref-training-preference-signal](https://datatracker.ietf.org/doc/draft-kelley-aipref-training-preference-signal/), which refuses training to every crawler that honors it.
### Layer 2: CDN-Native Bot Classification (0ms -- pre-computed)
@@ -277,7 +304,7 @@ The detection algorithm follows a strict priority order to minimize false positi
1. **Only process protected paths** -- never block requests to open content.
2. **Check for signed URL parameters first** -- if present, skip bot detection entirely and proceed to signature verification.
-3. **User-Agent matching is exact substring match**, not fuzzy. Unknown User-Agents pass through.
+3. **User-Agent matching uses the patterns in the lists**, case-insensitively, not fuzzy matching. A User-Agent that matches neither list passes through.
4. **Never block based on behavioral signals alone** -- they supplement User-Agent detection, not replace it.
## False Positive Handling
@@ -286,7 +313,7 @@ A false positive (regular user blocked as bot) is worse than a false negative (b
- **Only block on protected paths** -- never block requests to open content.
- **Never block requests with valid signed URLs** -- signature verification overrides bot detection.
-- **User-Agent matching is exact substring match**, not fuzzy. Unknown User-Agents pass through.
+- **User-Agent matching uses the patterns in the lists**, case-insensitively, not fuzzy matching. A User-Agent that matches neither list passes through.
- **Do not block based on behavioral signals alone** -- they supplement User-Agent detection, not replace it.
## Rate Limiting
diff --git a/website/src/content/docs/components/exchange/billing-adapter.mdx b/website/src/content/docs/components/exchange/billing-adapter.mdx
index d2a8a9e6..f90177d1 100644
--- a/website/src/content/docs/components/exchange/billing-adapter.mdx
+++ b/website/src/content/docs/components/exchange/billing-adapter.mdx
@@ -171,35 +171,35 @@ type DynamoDBBillingAdapter struct {
## Real Integration Example
-A Exchange operator (e.g., an existing ad-tech SSP adding resource access) would implement the interface against their existing billing system:
+An Exchange operator that already runs a billing system implements the interface against it:
```go
-// ExampleSSPBillingAdapter connects to SSP Corp's internal billing API.
-type ExampleSSPBillingAdapter struct {
+// ExampleBillingAdapter connects to the operator's internal billing API.
+type ExampleBillingAdapter struct {
httpClient *http.Client
baseURL string
apiKey string
}
-func (a *ExampleSSPBillingAdapter) Authorize(
+func (a *ExampleBillingAdapter) Authorize(
ctx context.Context, req *BillingAuthorizeRequest,
) (*BillingAuthorizeResponse, error) {
- // Call SSP Corp's internal billing API
- sspReq := &ssp.AuthorizeRequest{
+ // Call the operator's internal billing API
+ billReq := &billing.AuthorizeRequest{
AccountID: mapBillingRefToAccount(req.BillingRef),
Amount: req.Amount,
Currency: req.Currency,
Product: "content-licensing",
Reference: req.OfferID,
}
- sspResp, err := a.ssp.Authorize(ctx, sspReq)
+ billResp, err := a.billing.Authorize(ctx, billReq)
if err != nil {
- return nil, fmt.Errorf("ssp billing: %w", err)
+ return nil, fmt.Errorf("billing: %w", err)
}
return &BillingAuthorizeResponse{
- Approved: sspResp.Status == "approved",
- BillingID: sspResp.HoldID,
- Reason: sspResp.DeclineReason,
+ Approved: billResp.Status == "approved",
+ BillingID: billResp.HoldID,
+ Reason: billResp.DeclineReason,
}, nil
}
```
@@ -258,7 +258,7 @@ The estimate a metered offer carries (`Pricing.estimated_quantity`) is the publi
## Unit Cost Normalization
-DiscoverResources includes `unit_cost` in each offer so AI agents and Brokers can compare offers across pricing models: the price of one purchase divided by the resource's size in the comparison unit of its content type (tokens for text, seconds for audio and video, pages for documents, records for data). See [unit_cost: Normalized Pricing](/getting-started/how-money-flows/#unit_cost-normalized-pricing). The size is computed from content metadata. It is a comparison input only: it is not `estimated_quantity`, which is the publisher's estimate in the price's own unit, and it never enters a charge.
+DiscoverResources includes `unit_cost` in each offer so AI agents can compare offers across pricing models: the price of one purchase divided by the resource's size in the comparison unit of its content type (tokens for text, seconds for audio and video, pages for documents, records for data). See [unit_cost: Normalized Pricing](/getting-started/how-money-flows/#unit_cost-normalized-pricing). The size is computed from content metadata. It is a comparison input only: it is not `estimated_quantity`, which is the publisher's estimate in the price's own unit, and it never enters a charge.
### Size Formula
diff --git a/website/src/content/docs/components/exchange/configuration.mdx b/website/src/content/docs/components/exchange/configuration.mdx
index 3722149d..0f95b0e0 100644
--- a/website/src/content/docs/components/exchange/configuration.mdx
+++ b/website/src/content/docs/components/exchange/configuration.mdx
@@ -10,7 +10,7 @@ Three-tier configuration with precedence: environment variables > config file >
| Setting | Source | Hot Reload? | Example |
|---|---|---|---|
| Listen address | Env var | No (restart) | `EXCHANGE_LISTEN_ADDR=:8080` |
-| Exchange domain | Env var | No | `EXCHANGE_DOMAIN=exchange.ssp.com` |
+| Exchange domain | Env var | No | `EXCHANGE_DOMAIN=exchange.example` |
| TLS cert/key | Env var / file | No | `EXCHANGE_TLS_CERT=/etc/tls/cert.pem` |
| Tenant configs | Config file (YAML/JSON) | Yes (SIGHUP) | `EXCHANGE_CONFIG=/etc/exchange/config.yaml` |
| Signing keys | Secrets manager | Yes (polled) | `EXCHANGE_SECRETS_PROVIDER=aws-sm` |
@@ -26,12 +26,12 @@ Three-tier configuration with precedence: environment variables > config file >
```yaml
# /etc/exchange/config.yaml
exchange:
- domain: "exchange.ssp-example.com"
+ domain: "exchange.example"
version: "1.0"
billing:
adapter: "dynamodb" # or "http", "in-memory"
- endpoint: "https://billing.internal.ssp.com/v1"
+ endpoint: "https://billing.internal.example/v1"
timeout: "30ms"
timeout_policy: "reject" # or "defer"
diff --git a/website/src/content/docs/components/transaction-log/storage-backends.mdx b/website/src/content/docs/components/transaction-log/storage-backends.mdx
index 70c21502..7aeb8795 100644
--- a/website/src/content/docs/components/transaction-log/storage-backends.mdx
+++ b/website/src/content/docs/components/transaction-log/storage-backends.mdx
@@ -214,7 +214,7 @@ type TransactionEvent struct {
### Retention Policy
-13-month minimum retention for financial data. Matches ad-tech invoice cycles.
+13-month minimum retention for financial data.
| Data Class | Hot (fast query) | Warm (queryable, slower) | Cold (archive) | Total Retention |
|---|---|---|---|---|
diff --git a/website/src/content/docs/getting-started/for-ai-agents.mdx b/website/src/content/docs/getting-started/for-ai-agents.mdx
index 82353203..385df91d 100644
--- a/website/src/content/docs/getting-started/for-ai-agents.mdx
+++ b/website/src/content/docs/getting-started/for-ai-agents.mdx
@@ -291,7 +291,7 @@ Your agent connects and receives continuous data for the session duration. When
**Direct mode** (default): Your agent talks to each provider's Exchange directly. Simple. Works for known providers.
-**Brokered mode**: A FORA Broker queries multiple Exchanges in parallel, deduplicates offers (same article on different sites), filters by freshness and profile compatibility, and picks the cheapest or highest-quality source. Better for cost optimization at scale and cross-domain queries.
+**Brokered mode**: A FORA Broker queries multiple Exchanges in parallel, deduplicates offers (same article on different sites), filters by freshness and profile compatibility, and returns the offers in its order of relevance. Your agent decides which one to buy. Better for cross-domain queries, and for seeing every offer for the same content at once.
## Content Attestation
diff --git a/website/src/content/docs/getting-started/for-exchange-operators.mdx b/website/src/content/docs/getting-started/for-exchange-operators.mdx
index e7c3e75b..846f7bb8 100644
--- a/website/src/content/docs/getting-started/for-exchange-operators.mdx
+++ b/website/src/content/docs/getting-started/for-exchange-operators.mdx
@@ -1,146 +1,87 @@
---
title: "FORA for Exchange Operators"
-description: "How to operate a FORA Exchange — aggregate resource catalogs, handle transactions, earn a margin"
+description: "How to run a FORA Exchange on FORA Server: what you deploy, how publishers and agents join, the rules and fee you set, and the checks every Exchange runs."
---
-## The Opportunity
+An Exchange holds publishers' catalogs and terms, signs their offers, records each sale and takes usage reports. Anyone can run one. FORA Server, which is open source, includes the Exchange, and every FORA agent can buy from any FORA Exchange, because every Exchange speaks the same protocol. You set your own rules and your own fee.
-Resource providers need a transaction layer but don't want to build one. Courts don't want to run billing APIs. Publishers don't want to negotiate with every AI company. Hospitals don't want to handle metered access to imaging data. Patent offices don't want to build per-query pricing.
+Settlement stays outside the protocol: you bill the agents' account holders and pay your publishers through your own billing. FORA defines no payment rail.
-You solve this by running an **Exchange**: aggregate their catalogs, handle discovery, pricing, and transactions, and earn a margin on every access. This is the aggregator pattern — the same model that drives SSPs in ad-tech, clearinghouses in finance, and distributors in publishing.
+## Who runs an Exchange
-FORA was validated across **18 real-world scenarios** spanning 10 industries. The same protocol works for all of them:
+The same messages carry any catalog. What changes is the unit an offer is metered in and the domain profile its metadata follows:
-| Exchange Type | What You Aggregate | Providers | Example Operators |
-|---|---|---|---|
-| Media exchange | News articles, podcasts, video | Publishers | SSPs, media companies |
-| Academic exchange | Journal articles, preprints | Publishers, repositories | Elsevier, OpenAlex |
-| Legal exchange | Court filings, legislation, patents | Courts, patent offices | PACER aggregator, EUR-Lex |
-| Credit/financial exchange | Credit reports, market data | Data providers | D&B resellers, MarketData |
-| Medical data exchange | Clinical data, imaging | Hospitals, research institutions | TCIA, imaging exchanges |
-| Government data exchange | Property records, business filings | County/state agencies | CoreLogic, ATTOM |
-
-The value proposition is the same regardless of domain: providers get revenue from AI agents without building anything, agents get standardized access without bespoke integrations, and you keep a margin on every transaction.
-
-## What You Deploy
-
-You deploy an **Exchange Node** -- a Go service that handles three API calls:
-
-1. **DiscoverResources** -- AI agent asks "what's available at this URL and how much?" Your Exchange looks up the provider's catalog and returns offers with pricing.
-2. **ExecuteTransaction** -- AI agent says "I'll take it." Your Exchange checks the requester's balance, logs the transaction, and returns a signed URL.
-3. **ReportUsage** -- AI agent reports what it did with the content. Mandatory. You track compliance.
-
-That's the whole protocol. Three RPCs. Your billing adapter plugs in where `Authorize(requester, amount)` and `Record(transaction)` are called.
-
-## Unit-Agnostic Metering
-
-A media exchange meters in tokens. A legal exchange meters in pages. A medical exchange meters in studies. A geospatial exchange meters in sq_km. Same protocol, different units.
-
-FORA pricing states what is metered and at what price, with no assumption about the unit:
-
-- `unit`: what is metered.
-- `rate`: the price of one `unit`.
-- `estimated_quantity`: optional, the publisher's own estimate, in `unit`. A metered purchase charges the estimate × `rate`, or one unit's `rate` when the offer states no estimate, and that charge is final (see [Charging a metered purchase](/protocol/transaction-flow/#charging-a-metered-purchase)).
-- `unit_cost`: the comparison metric, the price of one purchase divided by the resource's size in the comparison unit of its content type. It never enters the charge.
-
-```
-Exchange Type Metering Unit Example rate
-──────────────────────────────────────────────────────────────
-Media tokens $0.00003/token
-Legal pages $0.50/page
-Academic accesses $0.05/access
-Medical imaging studies $2.00/study
-Credit/financial records $5.00/record
-Geospatial sq_km $0.10/sq_km
-Streaming (audio) minutes $0.02/minute
-Pharma interaction_pairs $0.001/pair
-```
-
-Usage reporting follows the same pattern: `consumed_quantity` + `consumed_unit`, which matches the offer's `unit`. A report records the quantity consumed and is never refused for it. Your Exchange MUST support these fields on all offers and MUST NOT assume an offer is metered in tokens; only a usage report that omits `consumed_unit` is read as tokens.
-
-## Extension Profiles
-
-Your Exchange declares which domain profiles it supports in its manifest. This tells Brokers and agents what kind of resources you handle:
-
-```json
-// Media exchange
-{ "supported_profiles": ["fora-news-v1"] }
-
-// Academic exchange
-{ "supported_profiles": ["fora-academic-v1"] }
-
-// Legal exchange
-{ "supported_profiles": ["fora-legal-v1"] }
+| Exchange over | What it holds, and from whom | Who might run it |
+|---|---|---|
+| News | Articles, podcasts and video, from publishers | A news publisher group, a press agency |
+| Scholarly content | Journal articles and preprints, from publishers and repositories | A university press, a society publisher |
+| Legal records | Court filings, legislation and patents, from courts and patent offices | A legal-data company |
+| Financial data | Credit reports and market data, from data providers | A data vendor |
+| Medical data | Clinical data and imaging, from hospitals and research institutions | A research consortium |
+| Public records | Property records and business filings, from agencies | A public-records company |
+| Music | Recordings and the works they record, from labels, publishers and collectives | A licensing collective |
-// Multi-domain exchange
-{ "supported_profiles": ["fora-news-v1", "fora-academic-v1"] }
-```
+An existing marketplace can run a FORA Exchange over its own catalog, beside the channels it already has. See [For Content Marketplaces](/getting-started/existing-marketplace/).
-Agents declare which profiles they understand, and a Broker forwards that list to your Exchange unchanged; it does not route by profile. When an academic research agent declares `fora-academic-v1` and you declare that profile, your offers carry its fields.
+## What you deploy
-Profile conformance means your Exchange sets the required `ext` fields defined in the profile spec and validates them against the profile's JSON Schema:
+FORA Server's Exchange is one Go binary in a container, `ghcr.io/fora-protocol/exchange`. It listens on two ports: the public Exchange listener, `8081` by default, and an internal admin listener, `8082`. To run it you provide:
-| Profile | Scope | What It Means for Your Exchange |
-|---|---|---|
-| `fora-news-v1` | News, podcasts, broadcasting | Set IPTC subject codes, correction status, podcast metadata in offer ext fields |
-| `fora-academic-v1` | Scholarly articles, preprints | Set DOI, OA status, version lifecycle in offer ext fields |
-| `fora-legal-v1` | Legislation, case law, patents | Set jurisdiction, amendment chains, ELI/ECLI in offer ext fields |
+- **Two PostgreSQL databases**: one for the catalog, the transaction log and the audit log (`EXCHANGE_DSN`), and one for the account registry (`EXCHANGE_SOR_DSN`).
+- **Redis** (`REDIS_URL`), which stops a signed request from being replayed. Every instance of a scaled-out Exchange shares one.
+- **An Ed25519 private key**, which signs offers and delivery URLs. An RSA key is needed only for a publisher whose CDN is AWS CloudFront.
+- **A ledger.** `FORA_BILLING_ADAPTER` chooses it: `tigerbeetle` keeps a prepaid balance per agent account in TigerBeetle, in the currencies `EXCHANGE_BILLING_LEDGER` lists. The default, `free`, approves every charge and records none.
+- **The Broker's address** (`EXCHANGE_BROKER_WELLKNOWN_URL`). The Exchange reads the Broker's withdrawn-key list from there and refuses to start without the address. See [Broker](/components/broker/overview/#who-runs-a-broker).
+- **Your hostname** (`EXCHANGE_DOMAIN`), and one seat row for it, which holds your fee and your agent policy.
-A Exchange MAY implement zero, one, or multiple profiles. You can start with one domain and expand.
+The catalog importer, `fora-import`, ships in the same image and runs as a job. The publisher console is a separate, optional service. You can deploy each component by hand, or with the Terraform modules in the FORA Server repository, which render a Docker Compose stack. The [Exchange deployment guide](https://github.com/FORA-Protocol/reference-implementation/blob/main/src/exchange/DEPLOYMENT.md) and the [runbook](https://github.com/FORA-Protocol/reference-implementation/blob/main/src/exchange/RUNBOOK.md) are in the FORA Server repository.
-## If You're an SSP: How FORA Maps to Your Infrastructure
+The Exchange answers these calls:
-If you operate an SSP or ad exchange, FORA maps directly to your existing concepts:
+| Call | What it does |
+|---|---|
+| `ExchangeService.Register` | Opens an agent's account and returns its `billing_ref` |
+| `ExchangeService.GetAccountStatus` | Returns the state of the agent's account |
+| `ExchangeService.DiscoverResources` | Returns signed offers for the URLs in a `ResourceQuery` |
+| `ExchangeService.ExecuteTransaction` | Buys offers and returns a signed delivery URL for each paid item |
+| `ExchangeService.ReportUsage` | Takes the agent's usage report and returns its `report_id` |
+| `CatalogService.PushResources` | Takes catalog entries from a contributor the site's `fora.json` names |
+| `AdminService.SetTenantFeeRate` | Sets your fee, on the admin listener only |
+| `AdminService.SetReportingPolicy` | Sets your reporting policy, on the admin listener only |
-```
-Ad-Tech Concept → FORA Equivalent
-─────────────────────────────────────────────
-SSP / Exchange → Exchange Node
-Publisher → Resource Provider
-ads.txt → fora.json (authorized providers)
-Bid Request → ResourceQuery
-Win Notification → TransactionRequest
-eCPM → unit_cost (effective cost per unit)
-Impression → Content access (signed URL)
-Viewability Report → Usage Report
-```
+It also serves `/.well-known/fora.json`, your key directory at `/.well-known/http-message-signatures-directory`, `/healthz` and `/readyz`.
-**You already have:**
-- Provider relationships and contracts
-- Billing infrastructure (invoicing, prepaid, credit)
-- CDN access for provider domains
-- Revenue share agreements
+## Your rules and your fee
-**FORA adds:**
-- A Exchange service (Go binary, stateless, horizontally scalable)
-- An edge function deployed to provider CDNs (bot blocking + signed URL verification)
-- A resource catalog imported from provider sitemaps, with the terms each provider sets in the publisher console
+- **Your fee.** Your commission is `fee_rate_bps` on your seat, in basis points, set with `AdminService.SetTenantFeeRate` without a restart. 1500 is 15%. Each sale records the gross, your fee and the publisher's net.
+- **Which agents can buy.** `activate_new_agents_by_default` decides whether a newly registered agent can buy at once, or waits until you activate its account. A deactivated account is refused paid purchases with `DENIAL_REASON_ACCOUNT_INACTIVE`.
+- **A welcome credit.** `EXCHANGE_DEFAULT_AGENT_CREDIT` grants each new account a starting balance, for example `EUR=100`.
+- **Purchases through a Broker.** `allow_broker_relay` on your seat decides whether you accept purchases the Broker relays. A refused item is denied with `DENIAL_REASON_RELAY_NOT_ACCEPTED`, and the agent can buy that offer directly.
+- **How many signatures a request may carry.** `EXCHANGE_MAX_INTERMEDIARY_HOPS`, published as `max_intermediary_hops`, 4 by default.
+- **Reporting.** `AdminService.SetReportingPolicy` sets which fields a usage report must carry. A report is never refused for the quantity it states.
-## Provider Onboarding
+The publishers set the terms and prices of their own content. Your Exchange signs the offers it builds from them.
-For each provider, you:
+## Publishers
-1. Generate their `fora.json` and have them publish it on their origin: the file they serve proves they control the domain and names your Exchange
-2. Import their pages from their sitemap into the catalog, with the terms they set in the publisher console, which fills them in from their RSL license when they publish one
-3. Configure their pricing tier
-4. Authorize `catalog_contributors` -- third-party verification vendors the provider permits to push attestations for their content
-5. If applicable, deploy the FORA edge function to their delivery endpoint (CDN, API gateway, PACS server) for signed URL verification
-6. Done. Their content is now available to AI requesters through your Exchange
+A publisher joins your Exchange through a publisher console:
-Providers publish that one file and deploy nothing else in the managed model. You manage the rest. This is the [Exchange-Managed deployment model](/architecture/deployment-models/). Some providers (large publishers, data vendors) may prefer to manage their own delivery endpoint -- the protocol supports both.
+1. It signs in and registers its business and its site.
+2. It sets its terms. The console fills them in from the site's RSL license when the site publishes one, and the publisher sets the price.
+3. The console generates the site's `fora.json`, which names your Exchange and lists your catalog contributor (`EXCHANGE_CATALOG_CONTRIBUTOR_HOST`) in `catalog_contributors`, and the edge function package for the site's CDN. The publisher serves the file and installs the package, usually in its own CDN account.
+4. The console checks both, and the listing goes live once the site proves it authorized your Exchange. `fora-import` then imports the pages of the site's sitemap into your catalog, with the publisher's terms, through `CatalogService.PushResources`.
-## Requester Onboarding
+From then on, the publisher sees its sales in the console. See [Publisher Onboarding](/getting-started/publisher-onboarding/) for the publisher's side.
-AI companies register via your existing account management or a self-signup API:
+## Agents
-- **Enterprise**: Out-of-band key exchange during contract signing. You register their Ed25519 public key.
-- **Self-signup**: AI agent registers via API with their public key manifest. Per-request billing only (no subscription).
+An agent opens an account at your Exchange with `ExchangeService.Register`. The Exchange identifies the agent by the key that signed the request and returns a `billing_ref`, the opaque, long-lived handle of the agent's account at your Exchange. Registering again returns the same handle. A paid purchase without an account is denied with `DENIAL_REASON_ACCOUNT_NOT_REGISTERED`; free content needs none.
-You don't need bilateral deals with every AI company. Any agent that speaks the FORA protocol can buy from your Exchange.
+You don't need a separate deal with each AI company. Any agent that speaks FORA can open an account and buy, under your agent policy. An agent on MCP registers through the [Identity Service](/components/identity/overview/).
-## Requester Verification
+## Requester verification
-The `Requester` message carries no signature: it states who is asking, and the RFC 9421 HTTP Message Signature on the request proves who sent it. Your Exchange MUST verify every request before acting on it. The rules are on the [Authentication](/protocol/authentication/) page; apply them in this order, and reject a request that fails any of them, with no fallback and no degraded mode:
+The `Requester` message carries no signature: it states who is asking, and the RFC 9421 HTTP Message Signature on the request proves who sent it. An Exchange MUST verify every request before acting on it, and FORA Server's Exchange does. The rules are on the [Authentication](/protocol/authentication/) page. An Exchange applies them in this order, and rejects a request that fails any of them, with no fallback and no degraded mode:
1. **Count the signatures first.** Before you verify any, refuse a request carrying more signatures than your `WellKnownManifest.max_intermediary_hops` with `resource_exhausted` (HTTP 429) and no typed reason. Every signature counts, whether or not it covers another. See [Verification at the Exchange](/protocol/authentication/#verification-at-the-exchange).
2. **Verify each signature under the Web Bot Auth profile.** Resolve its `keyid`, the key's RFC 7638 thumbprint, in the key directory its own covered `Signature-Agent` member names (`{origin}/.well-known/http-message-signatures-directory`), never from `Requester.domain`. Check the profile parameters, the four components a FORA RPC covers (`@method`, `@target-uri`, `content-digest`, `authorization`) and the `Content-Digest` of the body. Refuse a failure as `unauthenticated` with `request_auth_failure`: `SIGNATURE_MISSING` when no signature parses, `SIGNATURE_INVALID` when one does not verify, `SIGNATURE_STALE` outside its `created`/`expires` window. When a required component is missing, also answer with an `Accept-Signature` field listing the components you require, so the client can fix its request. See [Signature Verification](/protocol/authentication/#signature-verification) and [The Web Bot Auth Profile](/protocol/authentication/#the-web-bot-auth-profile).
@@ -152,9 +93,9 @@ The `Requester` message carries no signature: it states who is asking, and the R
Cache directories and keys with a TTL and re-fetch them to pick up rotation, without a fetch per request (see [Caching Contract](/protocol/authentication/#caching-contract)).
-### Delegation Verification
+### Delegation verification
-When a request includes a `Delegation` message (nested inside `Requester`), run the chain through the profile's verifier (`helpers.VerifyDelegation` in the Go SDK). The procedure, the limits and the checks are stated once, in [Verification](/protocol/authentication/#verification). This page adds only what your Exchange supplies to the verifier and what it does with the result:
+When a request includes a `Delegation` message (nested inside `Requester`), the Exchange runs the chain through the profile's verifier (`helpers.VerifyDelegation` in the Go SDK). The procedure, the limits and the checks are stated once, in [Verification](/protocol/authentication/#verification). This page adds only what the Exchange supplies to the verifier and what it does with the result:
1. **Trust comes from the subscription record.** Give the verifier the trusted root keys of the subscription record that owns the gated term, each with the issuer identity it belongs to. Trust is configured on that record, never fetched: not from `principal_domain`, not from `issuer`, not from a WBA directory, OIDC or JWKS. A key trusted on one record grants nothing on another. Withdraw trust by removing a key from the record's list; `revocation_uri` is not part of the profile.
2. **The holder is the key the request proved.** Give the verifier the thumbprint of that key: on a direct request the key that signed it (RFC 9421), on a purchase a Broker relays the key that signed each item's `AgentAcceptance`. The verifier refuses a mismatch. This is mandatory: it is what makes a stolen token unusable, so a token possessed without its signing key grants nothing. On a query a Broker originated, which proves possession of no agent key, do not verify or honor the delegation: answer as if none were present.
@@ -163,7 +104,7 @@ When a request includes a `Delegation` message (nested inside `Requester`), run
This verification is entirely offline — no callback to the principal is required.
-### Scope-Based Catalog Filtering
+### Scope-based catalog filtering
After verifying identity (and any delegation, as above), present the non-gated terms the requester's stated scopes cover; only those reach the `DiscoverResources` response. The scopes are the requester's own statement and you do not verify them. A gated term, a subscription term in the usual case, is different: present it only when the verified delegation's effective scopes cover it, ignore `requester.scopes` for that term, and answer a requester without a verified delegation as if the term did not exist. The matching rule is stated once, in [Scope Matching](/protocol/authentication/#scope-matching).
@@ -171,13 +112,13 @@ For a resource that exists but falls outside the requester's scopes, answer exac
An offer you presented stays purchasable until it expires, so a purchase is never refused because the requester's scopes do not cover the resource.
-### Subscription Quota Signaling
+### Subscription quota signaling
-When the requester holds a subscription (identifiable via the requester's `delegation` and/or subscription scopes), populate `SubscriptionQuotaInfo` entries on offers and transaction responses. This proactive quota signaling lets agents track remaining allowance and avoid committing to transactions they cannot complete. Track `max_accesses` and `quota_period` from the `Delegation` message when present.
+When the requester holds a metered subscription, an Exchange can report the allowance left in `SubscriptionQuotaInfo` entries on offers and transaction responses, so the agent does not commit to a purchase it cannot complete. An Exchange meters a subscription on its subscription record, never per delegation token, and names the record in `SubscriptionQuotaInfo.subscription_id`.
-### New Denial Reasons to Implement
+### Denial reasons
-FORA defines identity- and entitlement-related denial reasons your Exchange MUST return when appropriate. The delegation reasons:
+FORA defines identity- and entitlement-related denial reasons an Exchange MUST return when they apply. The delegation reasons:
| Denial Reason | When to Return |
|---|---|
@@ -203,121 +144,53 @@ And the relay reason, for a provider that does not accept purchases a Broker rel
This is the identity/entitlement subset with operator guidance; the complete, authoritative `DenialReason` enum is rendered from the proto on the [proto reference](/reference/proto-fora/). For the full identity model specification — Requester fields, the JWT chain, scope format, and key management — see [Authentication](/protocol/authentication/).
-## Attestation Verification
-
-Your Exchange verifies `ResourceAttestation` signatures at catalog push time. When a provider or verification vendor pushes resource metadata via `CatalogService.PushResources`, your Exchange:
-
-1. Fetches the signer's public key from their WBA directory (the JWK Set at `https://{verifier}/.well-known/http-message-signatures-directory`), identified by its RFC 7638 thumbprint (the RFC 9421 `keyid`)
-2. Verifies the Ed25519 signature over JCS-canonicalized claims
-3. Checks that the `verifier` is either the provider domain (Level 1 self-attestation) or a vendor listed in the provider's `catalog_contributors` (Level 2 third-party attestation)
-4. Rejects attestations from unauthorized contributors
-
-This prevents unauthorized parties from injecting false claims into your resource catalog.
-
-## Dispute Resolution
-
-Your Exchange handles `DisputeTransaction` RPCs from agents. How it resolves an accepted dispute is your policy; the recommended policy has three tiers. The volume and time figures are design goals, not measurements:
-
-| Tier | Scope | Volume (goal) | Response time (goal) |
-|---|---|---|---|
-| **Tier 1 -- Automated** | Delivery failures, content hash mismatches, signed URL expiry | ~99% of disputes | < 1 second |
-| **Tier 2 -- Rule-Based** | Ambiguous evidence, delivery says 200 but content is suspect | ~0.9% of disputes | < 24 hours |
-| **Tier 3 -- Investigation** | Aggregate fraud patterns, systematic abuse | ~0.1% of disputes | Case-by-case |
-
-**`report_id` generation**: When an agent calls `ReportUsage`, your Exchange returns a `report_id` in the `UsageReportResponse`. This `report_id` is required on every `DisputeRequest` -- it proves the agent fulfilled its reporting obligation before disputing. Your Exchange rejects disputes without a valid `report_id`.
-
-Delivery logs are the authoritative evidence source. They cannot be fabricated by the agent and provide independent verification of delivery status, response size, and timing.
-
-## Exchange Manifest and Health Checks
-
-As an Exchange operator, you are responsible for two additional capabilities:
-
-**Exchange Manifest** -- Your Exchange serves a machine-readable self-description at `/.well-known/fora.json` (`role=ROLE_EXCHANGE`). This tells Brokers and agents what your Exchange supports (pricing models, delivery methods, accepted verifiers, endpoints) without trial-and-error. Your offer-signing keys are published separately in your WBA directory (the JWK Set at `/.well-known/http-message-signatures-directory`), not in `fora.json`. The manifest includes:
+## Units and profiles
-- **`base_currency`** (ISO 4217 code) -- the currency for all `unit_cost` values. Enables cross-exchange price comparison by the Broker. A legal exchange quoting in EUR and a media exchange quoting in USD can be compared when both declare their currency.
-- **`supported_profiles`** -- domain extension profiles your Exchange conforms to.
+An offer names what it is metered in, and a metered purchase charges the publisher's estimate times the rate:
-The manifest is generated from your configuration at startup and served as a static JSON document. See [Exchange Manifest](/protocol/exchange-manifest/).
-
-**Health Checks** -- Your Exchange exposes three health layers: gRPC health (`grpc.health.v1.Health/Check`), HTTP liveness (`/healthz`), and HTTP readiness (`/readyz`). Brokers use gRPC health probing to make structured routing decisions -- they can distinguish an Exchange with a stale catalog from one with billing down. Kubernetes uses the HTTP endpoints for pod lifecycle management. See [Health Checks](/components/exchange/health-checks/).
-
-## Revenue Model
-
-Revenue depends on your exchange domain and the value of the resources you aggregate:
+- `unit`: what is metered.
+- `rate`: the price of one `unit`.
+- `estimated_quantity`: optional, the publisher's own estimate, in `unit`. A metered purchase charges the estimate × `rate`, or one unit's `rate` when the offer states no estimate, and that charge is final (see [Charging a metered purchase](/protocol/transaction-flow/#charging-a-metered-purchase)).
+- `unit_cost`: the price of one purchase divided by the resource's size in the comparison unit of its content type. It never enters the charge.
```
-Media exchange:
- Provider sets price: $0.05/article (tokens)
- AI Agent pays: $0.05/article
- You keep: Your standard rev-share margin
-
-Legal exchange:
- Provider sets price: $0.50/page (court filing)
- AI Agent pays: $0.50/page
- You keep: Your standard rev-share margin
-
-Credit exchange:
- Provider sets price: $5.00/record (credit report)
- AI Agent pays: $5.00/record
- You keep: Your standard rev-share margin
-
-Medical imaging exchange:
- Provider sets price: $2.00/study (MRI scan)
- AI Agent pays: $2.00/study
- You keep: Your standard rev-share margin
+Exchange over Pricing.unit Example rate
+──────────────────────────────────────────────────────────────
+News tokens $0.00003/token
+Legal records pages $0.50/page
+Scholarly content accesses $0.05/access
+Medical imaging images $2.00/image
+Financial data records $5.00/record
+Geospatial sq-km $0.10/sq-km
+Audio streams minutes $0.02/minute
+Drug interactions pharma:interaction_pairs $0.001/pair
```
-Higher-value resources mean higher absolute margins per transaction. A credit exchange processing 2K reports/day at $5.00 average = $10K/day gross = **$3.6M/year**. A media exchange processing 10K articles/day at $0.05 average = $500/day gross = **$182K/year**. Infrastructure costs scale with RPS, not transaction value -- approximately **$40/day** at 10K RPS.
+A usage report states `consumed_quantity` and `consumed_unit`, in the offer's `unit`, and is never refused for the quantity it reports. An Exchange MUST NOT assume an offer is metered in tokens; only a usage report that omits `consumed_unit` is read as tokens.
-## Multi-Exchange Competition
+Your manifest's `supported_profiles` tells Brokers and agents which domain profiles your offers follow. FORA Server's Exchange declares `fora-news-v1`, `fora-comp-v1` and `fora-music-v1`. An agent declares the profiles it understands, and a Broker forwards that list to your Exchange unchanged. See [Extension Profiles](/protocol/extension-profiles/).
-AI agents can use a **Broker** to query multiple Exchanges and compare prices -- like header bidding but for content. This means:
+## Exchange manifest and health
-- Competitive pressure keeps your margins honest
-- Premium content commands premium pricing (quality signals in offers)
-- Subscription deals give your providers an advantage (zero marginal cost for the requester)
-- `base_currency` in your manifest enables cross-exchange price comparison even across currencies
+Your Exchange serves its manifest at `/.well-known/fora.json`, with `role` set to `ROLE_EXCHANGE`. It tells Brokers and agents what your Exchange offers: its endpoint, `base_currency`, `supported_profiles` and `max_intermediary_hops`. In FORA Server, `base_currency` is the first currency in `EXCHANGE_BILLING_LEDGER`, and the Exchange rebuilds the manifest from its configuration. Your offer-signing keys are in your key directory, not in `fora.json`. See [Exchange Manifest](/protocol/exchange-manifest/).
-## Technical Integration
+`/healthz` checks the catalog database, and `/readyz` also checks the ledger. FORA Server's Broker reads the health endpoint of each Exchange in its registry and routes only to an Exchange that answers `200`.
-**NFR targets**: 10K RPS, under 100ms p99 latency, under $0.00005/transaction infrastructure cost.
+## Attestations
-**Billing adapter interface** (your integration point):
-```go
-type BillingAdapter interface {
- Authorize(ctx, billingRef, amount) → (ok, error)
- Record(ctx, transaction) → error
- CheckSubscription(ctx, billingRef, providerGroup) → (sub, error)
-}
-// billingRef is the account handle the Exchange minted at Register and
-// resolves on every call from what authenticates the agent: the request
-// signature on a direct request, the agent's acceptances on a purchase a
-// Broker relays. The account is the agent's, never the Broker's.
-```
-
-Implement these three methods against your existing billing system. The reference implementation uses DynamoDB; production deployments wire it to their real ledger.
-
-## Unit-Agnostic Pricing for Non-Text Resources
-
-FORA uses a single unit-agnostic pricing model, `PRICING_MODEL_PER_UNIT`, for all metered resources. `Pricing.unit` names what is metered, so non-text resources are priced through the same model:
-
-| Model | Unit | Use Case |
-|---|---|---|
-| `PRICING_MODEL_PER_UNIT` | `pages` | Court filings, legal documents, PDFs |
-| `PRICING_MODEL_PER_UNIT` | `minutes` | Video, audio, streaming sessions |
-| `PRICING_MODEL_PER_UNIT` | `records` | Database records, credit reports, medical studies |
+An attestation is a signed claim about a resource, by its publisher or by a verification vendor. When attestations arrive with `CatalogService.PushResources`, the protocol has the Exchange check that the signer is the provider itself or a contributor the provider's `catalog_contributors` lists, and verify the attestation's Ed25519 signature against the key in the signer's key directory. See [Resource Attestation](/protocol/content-attestation/).
-Pricing is unit-agnostic: `unit_cost` replaces eCPT, `estimated_quantity` replaces `estimated_tokens`, and `Pricing.unit` specifies the metering unit. Your Exchange MUST support these fields on all offers.
+## Usage reports and disputes
-## Streaming Delivery
+`ExchangeService.ReportUsage` takes a report on a transaction your Exchange recorded and returns a `report_id`. An agent files a dispute with `ExchangeService.DisputeTransaction` at the Exchange that sold the item, and each `DisputeRequest` names the `report_id` of its usage report. The protocol standardizes the dispute's states and outcomes; how an Exchange decides a dispute is up to the Exchange. See [Dispute Resolution](/protocol/dispute-resolution/).
-FORA supports `DELIVERY_METHOD_STREAMING` for real-time connections (WebSocket, SSE, HLS, Icecast). Your Exchange generates a signed URL pointing to a streaming endpoint. The agent connects and receives continuous data. Stream mechanics (heartbeat, reconnection, concurrent connection limits) are managed by the provider, not the protocol.
+## Brokers
-Streaming transactions follow the same FORA flow, but the signed URL has a longer TTL (streaming sessions last minutes/hours) and usage is reported in time-based units (minutes, hours).
+An agent can buy from your Exchange directly or through a [Broker](/components/broker/overview/). A Broker sends your Exchange only the URLs it routes to you, never the agent's words, and returns your signed offers unchanged, each as a full package, in its own order of relevance. The agent decides which to buy. Your account holder is always the agent: a purchase through a Broker charges the agent's account at your Exchange, never the Broker's.
-## Next Steps
+## Next steps
-1. Review the [Exchange design](/components/exchange/overview/) for architecture details
-2. See the [Billing Adapter](/components/exchange/billing-adapter/) for the integration interface
-3. Walk through the [Scenario Walkthrough](/protocol/scenario-walkthrough/) for a complete transaction trace
-4. Review [Authentication](/protocol/authentication/) for the signatures and keys every request, offer and delivery URL depends on
+1. Read the [Exchange deployment guide](https://github.com/FORA-Protocol/reference-implementation/blob/main/src/exchange/DEPLOYMENT.md) in the FORA Server repository.
+2. See [Publisher Onboarding](/getting-started/publisher-onboarding/) for what your publishers do.
+3. See [How Money Flows](/getting-started/how-money-flows/) for the billing chain and settlement.
+4. Review [Authentication](/protocol/authentication/) for the signatures and keys every request, offer and delivery URL depends on.
diff --git a/website/src/content/docs/getting-started/how-money-flows.mdx b/website/src/content/docs/getting-started/how-money-flows.mdx
index a33f7fb8..cfe0b889 100644
--- a/website/src/content/docs/getting-started/how-money-flows.mdx
+++ b/website/src/content/docs/getting-started/how-money-flows.mdx
@@ -5,7 +5,7 @@ description: "The FORA billing chain, pricing models, unit_cost normalization, a
## The Billing Chain
-FORA never touches money. It follows the OpenRTB model: the protocol handles authorization and logging, while actual settlement happens outside the protocol through existing billing infrastructure.
+FORA never touches money. The protocol handles authorization and logging, while actual settlement happens outside the protocol through existing billing infrastructure.
The chain works like this:
@@ -75,7 +75,7 @@ For example:
- Data at $0.002 per record (`unit: "records"`, `estimated_quantity: 500`): unit_cost = $0.002 per record, the rate itself
- Video at $0.50 per minute (`unit: "minutes"`, `estimated_quantity: 45`): unit_cost = $0.50 / 60 = $0.0083 per second
-The Broker uses `unit_cost` to compare offers across Exchanges and pricing models, selecting the most cost-effective option regardless of how the provider originally priced it.
+`unit_cost` lets an agent compare offers across Exchanges and pricing models by its own policy, however the provider priced them. A Broker does not order offers by it: it returns them in its order of relevance, and the agent decides.
Every `Offer` in a `ResourceResponse` includes both the native pricing and the computed `unit_cost` value. Here the article priced per access:
@@ -96,7 +96,7 @@ The purchase charges estimate × rate = 1 × $0.05, and the charge is final. Its
## base_currency
-Each Exchange declares a `base_currency` in its manifest (ISO 4217 code, e.g., "USD", "EUR"). All `unit_cost` values in offers from that Exchange are denominated in this currency. This enables cross-exchange price comparison by the Broker without requiring currency conversion at query time.
+Each Exchange declares a `base_currency` in its manifest (ISO 4217 code, e.g., "USD", "EUR"). All `unit_cost` values in offers from that Exchange are denominated in this currency. An agent can then compare offers from different Exchanges without converting currencies at query time.
## Subscription vs. Per-Request Billing
@@ -124,7 +124,7 @@ The AI company pays for a subscription off-protocol. On-protocol, a subscription
## Overspend Prevention
-Two enforcement layers prevent overspend, mirroring how OpenRTB prevents advertiser overspend:
+Two enforcement layers prevent overspend:
- **Agent/Broker-side**: `RequestConstraints.max_unit_cost` -- the agent will not accept offers above this threshold. `BrokerService.Resolve` may filter out offers that would exceed the agent's budget (`max_price`, and `period_budget` for its `budget_scope`) and answers a URI left with none with `OFFER_ABSENCE_REASON_BUDGET_EXCEEDED`. How the budget is computed is the Broker's own behaviour, not part of the protocol. Resolve never charges.
- **Exchange-side**: The Exchange operator's Billing Adapter checks the requester's balance or credit before authorizing each item. An item the requester cannot afford is denied in the response body (`DENIAL_REASON_INSUFFICIENT_BALANCE`), also when it is the only item.
@@ -137,27 +137,27 @@ Here is a concrete transaction with dollar amounts.
**Setup:**
- Provider: *The Daily Meridian*, a mid-size tech publisher
-- Exchange: SSP-Alpha, which sells The Daily Meridian's catalog
+- Exchange: Exchange Alpha, which sells The Daily Meridian's catalog
- Article: "AI Regulation: What Providers Need to Know" (2,500 words, ~3,300 tokens)
- The Daily Meridian's RSL license: $0.08 per crawl
-- SSP-Alpha's price for the article: $0.05
-- Requester: Acme AI, whose agent has a registered account at SSP-Alpha (account handle `RegisterResponse.billing_ref`, here `acct_7f3e21c9` — an opaque value minted by the Exchange at registration, meaningless on its own)
-- Acme AI has an annual subscription to The Daily Meridian through SSP-Alpha
+- Exchange Alpha's price for the article: $0.05
+- Requester: Acme AI, whose agent has a registered account at Exchange Alpha (account handle `RegisterResponse.billing_ref`, here `acct_7f3e21c9` — an opaque value minted by the Exchange at registration, meaningless on its own)
+- Acme AI has an annual subscription to The Daily Meridian through Exchange Alpha
**What happens:**
| Step | Action | Cost |
|---|---|---|
-| 1 | Acme AI's agent queries SSP-Alpha via `DiscoverResources`, signed with its own key and carrying its delegation from Acme AI | $0 |
-| 2 | SSP-Alpha returns two offers: per-access ($0.05) and subscription ($0.00) | -- |
+| 1 | Acme AI's agent queries Exchange Alpha via `DiscoverResources`, signed with its own key and carrying its delegation from Acme AI | $0 |
+| 2 | Exchange Alpha returns two offers: per-access ($0.05) and subscription ($0.00) | -- |
| 3 | The agent selects the subscription offer (zero marginal cost) | $0.00 |
-| 4 | `ExecuteTransaction` -- SSP-Alpha verifies the delegation chain; the price is zero, so no Billing Adapter call and no `billing_id` | $0.00 |
+| 4 | `ExecuteTransaction` -- Exchange Alpha verifies the delegation chain; the price is zero, so no Billing Adapter call and no `billing_id` | $0.00 |
| 5 | Signed URL issued, the agent fetches the article from CDN | -- |
| 6 | Usage report filed: 3,150 actual tokens consumed | -- |
**Financial attribution:**
- `cost.amount`: $0.00 (subscription)
-- The annual subscription fee is settled off-protocol between Acme AI and SSP-Alpha
+- The annual subscription fee is settled off-protocol between Acme AI and Exchange Alpha
**Without the subscription**, the per-request path would have charged $0.05 for this article, with a `unit_cost` of $0.00001515 per token.
@@ -178,7 +178,7 @@ The protocol does not prescribe how credits are settled between Exchange operato
## Pricing Transparency
-A Broker's fee has no field. It is additive and disclosed out of band, never folded into the offer price the Exchange signed, so the agent sees what it pays for the resource and what it pays for brokerage (see [Selection Engine](/components/broker/selection-engine/)).
+A Broker's fee has no field. It is additive and disclosed out of band, never folded into the offer price the Exchange signed, so the agent sees what it pays for the resource and what it pays for brokerage (see [Broker: order, budget and fees](/components/broker/overview/#order-budget-and-fees)).
## Reconciliation
@@ -188,7 +188,7 @@ Three independent records ensure transaction integrity:
2. **Exchange transaction log** (Exchange operator controls): every transaction with its `billing_id` (none for a zero-price item), unit cost and currency, the requester, the signed offer and the agent's signed acceptance, with the public keys their Ed25519 signatures verify under
3. **Usage reports** (agent files): what the agent actually did with the resource (token count, function, citation)
-All three records must agree. Discrepancies are resolved offline, like ad-tech billing discrepancies.
+All three records must agree. Discrepancies are resolved offline, between the parties.
A provider registered on the operator's publisher console sees its sales there. "Overview" totals transactions, gross, the Exchange fee and net for a period. "Transactions" lists each sale, filterable by domain and by usage-report status, and "Verify" checks the sale's two Ed25519 signatures in the browser. "Income" shows daily totals, with "Download CSV". The console is the operator's tool, not part of the protocol. See [After setup](/getting-started/publisher-onboarding/#after-setup) in Publisher Onboarding.
diff --git a/website/src/content/docs/getting-started/industry-landscape.mdx b/website/src/content/docs/getting-started/industry-landscape.mdx
deleted file mode 100644
index ec649f56..00000000
--- a/website/src/content/docs/getting-started/industry-landscape.mdx
+++ /dev/null
@@ -1,136 +0,0 @@
----
-title: "Industry Landscape"
-description: "The current state of AI metered resource access: standards ecosystem, closed platforms, and where FORA fits in the timeline."
----
-
-## The robots.txt Era Is Ending
-
-For decades, `robots.txt` was the only interface between web crawlers and providers. It offered a binary choice: allow or block. For search engines, this was sufficient -- crawling led to indexing, which led to traffic back to the provider.
-
-AI changes this equation. When an LLM ingests content for RAG, grounding, or training, the provider gets no return traffic. The content's value is extracted without compensation. Providers are responding by blocking AI crawlers wholesale, but this sacrifices the discovery benefits that AI-powered search and assistants could provide.
-
-The industry needs transactional infrastructure -- a way to say "yes, you can use this content, for this price, under these terms."
-
-## The Standards Ecosystem
-
-Three complementary standards are emerging to fill this gap:
-
-### RSL 1.0 (Really Simple Licensing)
-
-Released December 2025. Like `robots.txt` but for licensing.
-
-- Provider links a license file at any URL, for example from a `License:` line in `robots.txt`, or embeds the license in a page, feed or media file
-- Machine-readable terms: attribution requirements, pay-per-crawl, pay-per-inference, training restrictions
-- 1,500+ media organizations endorsing
-- **Limitation**: Declares terms but provides no transaction infrastructure. An agent can read what content costs but has no way to pay for it through RSL alone.
-
-### IAB Tech Lab CoMP v1.0 (Content Monetization Protocol)
-
-Released March 2026, in public comment through April 2026.
-
-- Defines the **data model** for resource access: Package, AISystem, Retrieval, Scope objects
-- Provides a logging component for audit and invoicing
-- Backed by IAB Tech Lab (Prebid's parent organization) -- carries significant weight with SSPs and exchanges
-- **Limitation**: Deliberately leaves out billing, settlement, and transport. Defines what to describe but not how to transact.
-
-### FORA (Federated Open Resource Access)
-
-The open transaction protocol that completes the stack.
-
-- Adds discovery (`fora.json`), transaction execution, signed URL delivery, and usage reporting
-- Resource attestation (three verification levels) and automated dispute resolution
-- Provider onboarding where the `fora.json` a domain serves proves control of that domain
-- Layers on RSL (the publisher console reads a provider's RSL license to fill in its terms) and CoMP (wraps CoMP objects in transaction messages)
-- Multi-exchange architecture with an optional Broker for cross-exchange comparison
-
-## Closed Platforms
-
-Two proprietary approaches are operating today:
-
-### Microsoft Provider Content Exchange (PCM)
-
-Launched February 2026.
-
-- **Walled garden** -- Microsoft operates the exchange exclusively
-- Distinguishes **grounding** (per-request, real-time content for Copilot/Bing) from **training** (bulk licensing for model training)
-- "Pay per demonstrated value" -- providers receive compensation based on content performance metrics
-- Early partners: Associated Press, Vox Media, Conde Nast, Hearst, Business Insider
-- **Limitation**: Only works within Microsoft's ecosystem. A provider licensing through PCM cannot serve the same content to non-Microsoft AI systems through the same infrastructure.
-
-### TollBit
-
-Startup with $7M in funding, operational today.
-
-- Running per-request content licensing in production
-- Per-1000-pages pricing, providers set their own rates
-- Charges AI customers a transaction fee but does not take a provider revenue share
-- **Limitation**: Proprietary API. Providers and AI companies are locked into TollBit's infrastructure. No interoperability with other licensing systems.
-
-## Why Open Standards Matter
-
-The ad-tech industry learned this lesson two decades ago. Before OpenRTB, every ad exchange had its own proprietary protocol. Advertisers and providers had to build custom integrations for each exchange, creating lock-in and fragmentation.
-
-OpenRTB changed this by defining a common protocol that any exchange, DSP, or SSP could implement. The result: a liquid market where requesters and providers could transact across any compliant exchange.
-
-AI resource access is at the same inflection point. Microsoft PCM and TollBit are building walled gardens -- the equivalent of pre-OpenRTB proprietary exchanges. FORA aims to be the OpenRTB for resource access: an open protocol that any Exchange, AI company, or provider can implement.
-
-## Comparison
-
-| Layer | RSL | CoMP | Microsoft PCM | TollBit | FORA |
-|---|---|---|---|---|---|
-| Terms declaration | RSL license | Package object | Provider dashboard | Provider dashboard | Publisher console, filled in from the RSL license |
-| Discovery | `robots.txt`, `Link` header, HTML, RSS or file metadata | Not defined | Microsoft index | TollBit API | `fora.json` + edge function fallback |
-| Transaction protocol | None | Not defined | Proprietary | Proprietary | **Open protocol** |
-| Content verification | None | None | Not defined | Not defined | **Three-level attestation** |
-| Dispute resolution | None | None | Not defined | Not defined | **Automated (three-tier)** |
-| Billing/settlement | None | Out of scope | Microsoft handles | TollBit handles | Billing adapter (pluggable) |
-| Multi-exchange | No | Implied | No (Microsoft only) | No (TollBit only) | **Yes (Broker)** |
-| Domain extension profiles | None | None | None | None | **Three profiles (news, academic, legal)** |
-
-## Three-Layer Governance
-
-FORA v1.0 prevents protocol fragmentation through a three-layer governance model:
-
-| Layer | Scope | Governance | Example |
-|---|---|---|---|
-| **1. Core Proto** | Fields in `fora.proto` | Protocol version bump | `resource_mutability`, `data_as_of` |
-| **2. Standard Extensions** | `fora.*` namespace in ext | Defined in FORA spec | `fora.compliance.dua_required` |
-| **3. Domain Profiles** | Domain namespace in ext | Published by domain communities | `news.iptc_guid`, `academic.doi` |
-
-Three domain extension profiles ship with v1.0: `fora-news-v1` (news, podcasts, broadcasting), `fora-academic-v1` (scholarly articles, preprints, datasets), and `fora-legal-v1` (legislation, case law, patents). Each profile is grounded in existing industry standards (IPTC NewsML-G2, CrossRef/OpenAlex, ELI/ECLI/Akoma Ntoso).
-
-Exchanges declare conformance via `supported_profiles` in their manifest. Agents declare which profiles they understand in the request. A Broker forwards the agent's declaration to each Exchange it queries and does not route by it; an Exchange that supports `fora-legal-v1` answers a legal research agent with that profile's fields.
-
-## Scope Boundary
-
-FORA v1.0 draws a clear boundary: **resource access is in scope; job execution is not.**
-
-The test: "Does the resource exist before the transaction?" If yes, it is a FORA use case. If the transaction creates something new from agent input (translation, ML inference, document generation), it is not.
-
-**In scope:** static resources (articles, papers, patents), dynamic resources (credit reports, drug databases), streaming resources (quote feeds, live broadcasts), database lookups (drug interaction queries).
-
-**Out of scope:** translation, ML inference, document generation. A Exchange offering these should use a different protocol. However, a pre-computed translation database (translations as resources) is a valid FORA use case.
-
-## Timeline
-
-| Date | Event |
-|---|---|
-| December 2025 | RSL 1.0 released, 1,500+ providers endorsing |
-| February 2026 | Microsoft PCM launched with initial provider partners |
-| March 2026 | IAB CoMP v1.0 released for public comment |
-| March 2026 | FORA v1.0: metering generalization, streaming delivery, content mutability, extension profiles, CoMP alignment |
-| April 2026 | IAB CoMP public comment period closes |
-
-## The Window for Open Standards
-
-Three forces create urgency:
-
-1. **IAB CoMP is being defined now** -- the public comment period closes April 2026. The industry is actively shaping the data model. FORA builds on CoMP, so alignment matters.
-2. **Walled gardens are growing** -- Microsoft and TollBit are signing providers to exclusive or preferred deals. Each provider locked into a proprietary system is one fewer participant in an open market.
-3. **AI traffic doubles quarterly** -- every month without monetization infrastructure represents lost revenue for providers and growing legal risk for AI companies operating without clear licensing.
-
-## Next Steps
-
-- [What Is FORA?](/getting-started/what-is-fora/) -- protocol overview and how it works
-- [Standards Layering](/protocol/standards-layering/) -- technical detail on how RSL, CoMP, and FORA interoperate
-- [Discovery Paths](/protocol/discovery-paths/) -- how agents find and enter the FORA flow
diff --git a/website/src/content/docs/getting-started/publisher-onboarding.mdx b/website/src/content/docs/getting-started/publisher-onboarding.mdx
index 232242ca..1e8f6739 100644
--- a/website/src/content/docs/getting-started/publisher-onboarding.mdx
+++ b/website/src/content/docs/getting-started/publisher-onboarding.mdx
@@ -8,7 +8,7 @@ import { FORA_PUBLISHER_CONSOLE_URL } from 'astro:env/client';
The publisher console takes your site from registration to its first sale. You name your business and tell it your domain and your terms, and it generates the two things that run on your side: a discovery file (`fora.json`) and an edge package for your CDN. You serve one file and install one package. From then on, AI agents pay for your pages and your readers are unaffected.
-You generate no keys and you never push a catalog. The operator, the company that runs the console and the Exchange, imports your pages from your sitemap. Registering costs nothing. Start at the publisher console.
+You generate no keys and you never push a catalog. The operator, the company that runs the console and the Exchange, imports your pages from your sitemap. FORA is open-source software, not an operator: any company can run FORA Server. The console on this site is run by Postindustria, which charges publishers no Exchange fee; only the payment system's transaction fees apply. Registering costs nothing. Start at the publisher console.
## Before you start
diff --git a/website/src/content/docs/getting-started/what-is-fora.mdx b/website/src/content/docs/getting-started/what-is-fora.mdx
index c331173f..0706bf63 100644
--- a/website/src/content/docs/getting-started/what-is-fora.mdx
+++ b/website/src/content/docs/getting-started/what-is-fora.mdx
@@ -25,7 +25,7 @@ An agent buys access to a resource from an Exchange, directly or through an opti
**For resource providers:** Expose your catalog through an Exchange. Set your price. Get paid per access. No custom integrations per consumer.
-**For Exchange operators:** Aggregate provider catalogs. Handle transactions. Earn a margin. Compete on coverage, quality, and price.
+**For Exchange operators:** Run an Exchange on FORA Server, which is open source. Hold your providers' catalogs and terms, sign their offers and record each sale, under your own rules and fees. Every FORA agent can open an account and buy from it.
## How It Works
@@ -155,15 +155,16 @@ FORA covers **metered access to resources that exist**. The test: "Does the reso
| Streaming resources (quote feeds, live audio) | Compute services |
| Database lookups (drug interactions, property records) | |
-## What Exists Today
+## What FORA Is Made Of
- **The wire format**: the `fora.v1` and `fora.admin.v1` protobuf packages, with a [JSON Schema](/reference/json-schemas/) for each message. See [Proto: FORA v1](/reference/proto-fora/).
- **SDKs in three languages**: {PACKAGES.go.coordinate} for Go, {PACKAGES.ts.coordinate} on npm and {PACKAGES.python.coordinate} on PyPI.
- **Six extension profiles**: [news](/protocol/ext-news/), [academic](/protocol/ext-academic/), [legal](/protocol/ext-legal/), [CoMP](/protocol/ext-comp/), [C2PA](/protocol/ext-c2pa/) and [music](/protocol/ext-music/).
-- **Two live demos** on a demo Exchange. In [First License](/getting-started/poc-walkthrough/), an agent buys a license to an article whose site refuses its direct request. In [Music](/getting-started/live-demo-music/), an agent buys two different licenses to the same track.
-- **FORA Server**: an [open-source implementation](https://github.com/FORA-Protocol/reference-implementation) of the Exchange, the Broker, an identity service with an MCP adapter, and the edge function.
+- **FORA Server**: an [open-source implementation](https://github.com/FORA-Protocol/reference-implementation) of the Exchange, the Broker, the Identity Service with its MCP endpoint, and the edge function. Anyone can run it.
- **The standards it builds on**: IAB Tech Lab's CoMP V1 is mapped field for field and carried as the `fora-comp-v1` profile. The RSL 1.0 AI-use terms seed FORA's restriction vocabulary. See [Standards Layering](/protocol/standards-layering/).
+Two live demos run on a demo Exchange. In [First License](/getting-started/poc-walkthrough/), an agent buys a license to an article whose site refuses its direct request. In [Music](/getting-started/live-demo-music/), an agent buys two different licenses to the same track.
+
## Next Steps
- [For Resource Providers](/getting-started/for-providers/) — how to expose your resources through FORA
diff --git a/website/src/content/docs/index.mdx b/website/src/content/docs/index.mdx
index 367e54dd..40d3da44 100644
--- a/website/src/content/docs/index.mdx
+++ b/website/src/content/docs/index.mdx
@@ -38,7 +38,7 @@ import { Card, CardGrid, LinkCard } from '@astrojs/starlight/components';
Where an operator runs a publisher console, name your business, add your domain and connect your site: one generated file on your site and one package on your CDN. No keys to generate, no catalog to push.
- No walled gardens. Exchange operators, providers, and third parties operate Exchanges. A Broker compares offers across them.
+ Anyone can run an Exchange: an operator, a provider or a third party. A Broker brings their offers to the agent in one answer.
Zero-SDK resource access. Add the FORA MCP server to any MCP-capable agent and discover, transact, and dispute resources without custom integration.
@@ -49,7 +49,7 @@ import { Card, CardGrid, LinkCard } from '@astrojs/starlight/components';
- [What is FORA?](/getting-started/what-is-fora/) -- [How Money Flows](/getting-started/how-money-flows/) -- [Industry Landscape](/getting-started/industry-landscape/)
+ [What is FORA?](/getting-started/what-is-fora/) -- [How Money Flows](/getting-started/how-money-flows/) -- [Content marketplaces](/resources/content-marketplaces/) -- [The standards map](/resources/standards-map/)
[Transaction Flow](/protocol/transaction-flow/) -- [Discovery Paths](/protocol/discovery-paths/) -- [Authentication](/protocol/authentication/)
diff --git a/website/src/content/docs/protocol/discovery-paths.mdx b/website/src/content/docs/protocol/discovery-paths.mdx
index 8e62bc48..aa390275 100644
--- a/website/src/content/docs/protocol/discovery-paths.mdx
+++ b/website/src/content/docs/protocol/discovery-paths.mdx
@@ -29,12 +29,12 @@ sequenceDiagram
Agent->>PUB: GET /.well-known/fora.json
PUB-->>Agent: WellKnownManifest (role=ROLE_PUBLISHER, exchange endpoints)
- Note over Agent: Content at /premium/* requires licensed access via exchange.ssp.com
+ Note over Agent: Content at /premium/* requires licensed access via exchange.example
Agent->>MP: DiscoverResources (ResourceQuery)
MP-->>Agent: ResourceResponse (Offer[] with pricing)
- Note over Agent: Select best offer Check budget constraints
+ Note over Agent: Choose an offer Check budget constraints
Agent->>MP: ExecuteTransaction (TransactionRequest)
MP-->>Agent: TransactionResponse (signed URL + billing_id)
@@ -63,7 +63,7 @@ sequenceDiagram
## Path B: Broker-Mediated Multi-Exchange
-The agent hands discovery to a FORA Broker, which chooses the Exchanges, queries multiple Exchanges in parallel, compares offers, and selects the best deal.
+The agent hands discovery to a FORA Broker, which chooses the Exchanges, queries them in parallel, and returns their signed offers in its order of relevance. The agent decides which offer it buys.
```mermaid
sequenceDiagram
@@ -83,13 +83,13 @@ sequenceDiagram
MPB-->>Orch: ResourceResponse (offers)
end
- Note over Orch: Compare offers by unit_cost Deduplicate by ResourceIdentity Apply budget constraints Rank offers per URI
+ Note over Orch: Group by ResourceIdentity Apply the agent's constraints Order offers per URI by relevance
Orch-->>Agent: DiscoveryResponse (offer_groups: signed offers per URI)
- Note over Agent: Select best offer (each Offer carries Offer.exchange, the execute-routing target)
+ Note over Agent: Choose an offer (each Offer carries Offer.exchange, the execute-routing target)
- Agent->>MPA: ExecuteTransaction (winning offer)
+ Agent->>MPA: ExecuteTransaction (chosen offer)
MPA-->>Agent: TransactionResponse (signed URL + cost + billing_id)
Agent->>CDN: GET signed URL
@@ -104,12 +104,12 @@ sequenceDiagram
|---|---|---|
| Agent to Broker | `DiscoveryRequest` with `RequestConstraints` | **FORA** (budget, authorized exchanges) |
| Broker to Exchanges | `ResourceQuery` (parallel fanout) | **FORA** wrapping **FORA Requester** |
-| Offer comparison | `Offer.pricing.unit_cost` | **FORA** unit cost normalization |
+| Offer comparison, by the agent | `Offer.pricing.unit_cost` | **FORA** unit cost normalization |
| Content dedup | `Offer.identity` (canonical_url, content_hash) | **FORA** ResourceIdentity extension |
| Content attestations | `Offer.attestations` | **FORA** v1.0 `ResourceAttestation` |
| Broker to Agent | `DiscoveryResponse` (offer_groups: signed offers per URI) | **FORA** -- discovery only, no delivery |
-| Execute winning offer | `ExecuteTransaction` / `TransactionResponse` (signed URL + cost + billing_id) | **FORA** (separate execute step, routed via `Offer.exchange`) |
-| Usage reporting | `UsageReport` to the winning Exchange | **FORA** |
+| Execute the chosen offer | `ExecuteTransaction` / `TransactionResponse` (signed URL + cost + billing_id) | **FORA** (separate execute step, routed via `Offer.exchange`) |
+| Usage reporting | `UsageReport` to the Exchange that issued the offer | **FORA** |
## Path C: Accidental Discovery via Edge Function (403)
@@ -240,7 +240,7 @@ sequenceDiagram
## Path E: Batch Multi-URL Acquisition
-The agent needs multiple resources in a single request (e.g., RAG grounding across several sources). The Broker sends all URIs in one `ResourceQuery` to each Exchange, receives `OfferGroups`, and returns the ranked signed offers per URI in a single `DiscoveryResponse`. Discovery stops there: the agent selects the best offer per URI and commits the selections itself via batch `ExecuteTransaction` calls (one per Exchange, routed by `Offer.exchange`), each returning a `TransactionResponse`.
+The agent needs multiple resources in a single request (e.g., RAG grounding across several sources). The Broker sends all URIs in one `ResourceQuery` to each Exchange, receives `OfferGroups`, and returns the signed offers per URI, in its order of relevance, in a single `DiscoveryResponse`. Discovery stops there: the agent chooses an offer per URI and commits the selections itself via batch `ExecuteTransaction` calls (one per Exchange, routed by `Offer.exchange`), each returning a `TransactionResponse`.
```mermaid
sequenceDiagram
@@ -260,11 +260,11 @@ sequenceDiagram
MPB-->>Orch: ResourceResponse (3 OfferGroups)
end
- Note over Orch: Per URI: merge offers across MPs, deduplicate, rank
+ Note over Orch: Per URI: merge offers across MPs, group, order by relevance
Orch-->>Agent: DiscoveryResponse (offer_groups: signed offers per URI)
- Note over Agent: Select best offer per URI Sum costs across URIs -> check budget Group selections by Offer.exchange
+ Note over Agent: Choose an offer per URI Sum costs across URIs -> check budget Group selections by Offer.exchange
Agent->>MPA: ExecuteTransaction (batch: items[] with 2 offers)
MPA-->>Agent: TransactionResponse (items[] + total_cost)
@@ -365,7 +365,7 @@ sequenceDiagram
- **Subscription detection at query time**: The Exchange checks the requester's subscription status and includes offers with `subscription_id` and `rate=0`.
- **Reporting is still mandatory**: Subscription offers carry `ReportingObligation.required = true`. Providers need consumption data for subscription renewals and content strategy.
- **Quota enforcement**: The Exchange checks remaining quota before authorizing subscription transactions. If the subscription's token quota is exhausted, the subscription offer is not included.
-- **Broker preference**: The Broker's selection engine ranks subscription offers above all per-request offers (zero marginal cost). When `preferred_exchanges` is set and the agent has subscriptions there, those offers get top priority.
+- **Broker order**: The agent names the Exchanges it holds subscriptions with in `RequestConstraints.preferred_exchanges`, and the Broker SHOULD put their offers first when they have the resource. The subscription offer shows its zero marginal cost in its own pricing, and the agent decides whether to take it.
- **Financial attribution**: The `subscription_unit_value` field carries the per-unit cost even though `cost.amount` is 0, enabling prepaid drawdown accounting.
## Next Steps
diff --git a/website/src/content/docs/protocol/ext-news.mdx b/website/src/content/docs/protocol/ext-news.mdx
index 9897d32e..80403ee1 100644
--- a/website/src/content/docs/protocol/ext-news.mdx
+++ b/website/src/content/docs/protocol/ext-news.mdx
@@ -580,7 +580,7 @@ The Exchange returns an `OfferGroup` per URI. The Broker can then:
1. Match offers across Exchanges by `news.iptc_guid` (if the Reuters and AP stories syndicated through wire)
2. Filter out any offers where `news.pub_status` is not `"usable"`
3. Prefer offers with higher `news.version` (latest correction)
-4. Compare `unit_cost` across Exchanges for the same underlying story
+4. Return every remaining offer for the same underlying story as one group, in its order of relevance; the agent chooses which to buy
---
diff --git a/website/src/content/docs/protocol/role-composition.mdx b/website/src/content/docs/protocol/role-composition.mdx
index 5a612a65..adab81eb 100644
--- a/website/src/content/docs/protocol/role-composition.mdx
+++ b/website/src/content/docs/protocol/role-composition.mdx
@@ -10,7 +10,7 @@ FORA defines four roles via the `Role` enum in `WellKnownManifest`:
| Role | Responsibility |
|---|---|
| `ROLE_AGENT` | Issues `ResourceQuery` / `TransactionRequest` / `UsageReport`; signs requests with Ed25519 |
-| `ROLE_BROKER` | Fans out across Exchanges, compares offers, deduplicates by `ResourceIdentity`, enforces budget constraints |
+| `ROLE_BROKER` | Fans out across Exchanges, groups offers by `ResourceIdentity`, returns them in its order of relevance, applies the agent's budget constraints |
| `ROLE_EXCHANGE` | Holds the resource catalog and pricing, authorises transactions, issues signed retrieval endpoints |
| `ROLE_PUBLISHER` | Owns the content, declares RSL terms, optionally self-attests resources, serves delivery |
@@ -42,7 +42,7 @@ In other words: the protocol does not assume that distinct roles are run by dist
## Implications
-- An operator that runs both an Exchange and a Broker MUST NOT route every Broker request to its own Exchange when other Exchanges hold matching offers at lower `unit_cost`. The protocol does not enforce this behaviourally, but the Broker's offer comparison is auditable from its outbound signed `ResourceQuery` traffic and the resulting Offers — agents and regulators can verify routing fairness from the wire.
+- An operator that runs both an Exchange and a Broker MUST NOT route every Broker request to its own Exchange when other Exchanges hold matching offers. The agent gets the other Exchanges' offers too, in the Broker's order of relevance, and decides. The protocol does not enforce this behaviourally, but the Broker's routing is auditable from its outbound signed `ResourceQuery` traffic and the resulting Offers — agents and regulators can verify routing fairness from the wire.
- An operator that runs both an Agent and an Exchange MUST sign the Agent's `TransactionRequest` with the Agent role's key, not the Exchange's. Mixing keys is an attestation forgery; the resulting offer trail will fail verification by anyone who pulls the relevant `WellKnownManifest` files.
- An operator that runs a Publisher and an Exchange MAY self-attest resources at Level 1 (publisher signs its own `ResourceAttestation`). Agents that require Level 2 verification will discount or ignore Level 1 self-attestations from the same operator that runs the Exchange — see [Content Attestation](/protocol/content-attestation/#three-verification-levels) for trust-policy patterns.
diff --git a/website/src/content/docs/protocol/scenario-walkthrough.mdx b/website/src/content/docs/protocol/scenario-walkthrough.mdx
index d4e152d9..bc3bd454 100644
--- a/website/src/content/docs/protocol/scenario-walkthrough.mdx
+++ b/website/src/content/docs/protocol/scenario-walkthrough.mdx
@@ -9,8 +9,8 @@ description: "End-to-end FORA v1.0 walkthrough: provider setup, agent request, s
**Agent**: Anthropic's Claude (`claude.ai`), running the FORA Agent SDK
**Broker**: Anthropic's hosted Broker (`broker.anthropic.com`)
-**Subscription**: Anthropic has an annual deal with Hearst Media via SSP-Alpha exchange
-**Request**: Claude needs an article about AI regulation originally published by Hearst's TechCrunch (`techcrunch.com`), syndicated to The Verge (`theverge.com`) via SSP-Beta exchange
+**Subscription**: Anthropic has an annual deal with Hearst Media via Exchange Alpha
+**Request**: Claude needs an article about AI regulation originally published by Hearst's TechCrunch (`techcrunch.com`), syndicated to The Verge (`theverge.com`) via Exchange Beta
**Extension Profiles**: Both Exchanges declare `supported_profiles: ["fora-news-v1"]` in their manifests. The Broker forwards `supported_profiles` from the agent to enable profile-aware metadata in offers (e.g., `news.iptc_guid`, `news.correction_status`).
---
@@ -48,8 +48,8 @@ Before any agent request, providers configure their domains and Exchanges ingest
"contact": "licensing@hearst.com",
"exchanges": [
{
- "domain": "exchange.ssp-alpha.com",
- "endpoint": "https://exchange.ssp-alpha.com/fora/v1",
+ "domain": "exchange.alpha.example",
+ "endpoint": "https://exchange.alpha.example/fora/v1",
"relationship": "PROVIDER_RELATIONSHIP_DIRECT"
}
]
@@ -114,15 +114,15 @@ The Broker announces its key in its own WBA directory the same way:
}
```
-### 0c. SSP-Alpha Resource Ingestion
+### 0c. Exchange Alpha Resource Ingestion
-SSP-Alpha's Resource Ingestion Pipeline runs in the background:
+Exchange Alpha's Resource Ingestion Pipeline runs in the background:
```
1. Crawl techcrunch.com/sitemap.xml -> discover all /premium/* URLs
2. Fetch the RSL license robots.txt names -> extract permits/prohibits and its rate ($0.08/crawl), Hearst's cap
3. HTML crawl via readability -> extract text -> word count -> estimate tokens
-4. Apply Hearst's private pricing (from SSP-Alpha config DB): $0.05/article
+4. Apply Hearst's private pricing (from Exchange Alpha config DB): $0.05/article
(Below the RSL ceiling of $0.08 — compliant)
5. Third-party attestation (GumGum via CatalogService.PushResources):
GumGum crawls the article, signs attestation claims with Ed25519:
@@ -168,31 +168,31 @@ SSP-Alpha's Resource Ingestion Pipeline runs in the background:
7. Compile into radix trie -> serialize to catalog.bin -> Exchange loads via atomic pointer swap
```
-SSP-Alpha also knows about Anthropic's subscription, looked up by Anthropic's account handle (minted at registration):
+Exchange Alpha also knows about Anthropic's subscription, looked up by Anthropic's account handle (minted at registration):
```
BillingAdapter.CheckSubscription("acct_7f3e21c9", "hearst-media")
-> { subscription_id: "SUB-ANTHROPIC-HEARST-2026", active: true, quota_remaining: 850000 }
```
-### 0d. The Verge Syndication via SSP-Beta
+### 0d. The Verge Syndication via Exchange Beta
```
1. The Verge reprinted the TechCrunch article (syndication deal):
theverge.com/premium/ai-regulation-republish.html
-2. The Verge has fora.json pointing to SSP-Beta:
+2. The Verge has fora.json pointing to Exchange Beta:
{
"ver": "1.0",
"provider": "theverge.com",
"exchanges": [{
- "domain": "exchange.ssp-beta.com",
- "endpoint": "https://exchange.ssp-beta.com/fora/v1",
+ "domain": "exchange.beta.example",
+ "endpoint": "https://exchange.beta.example/fora/v1",
"relationship": "PROVIDER_RELATIONSHIP_DIRECT"
}]
}
-3. SSP-Beta's ingestion pipeline:
+3. Exchange Beta's ingestion pipeline:
- Discovers the article, estimates 3100 tokens
- The Verge's pricing: $0.07/article (higher than TechCrunch, below RSL ceiling)
- Identity: same iptc_guid (syndicated from same TechCrunch wire)
@@ -204,7 +204,7 @@ BillingAdapter.CheckSubscription("acct_7f3e21c9", "hearst-media")
resource_mutability: RESOURCE_MUTABILITY_STATIC
}
- Attestations: [{ verifier: "gumgum.com", claims: { estimated_quantity: 3100, word_count: 2350 } }]
- - Anthropic has NO subscription with SSP-Beta
+ - Anthropic has NO subscription with Exchange Beta
```
---
@@ -227,8 +227,8 @@ Step 1: Parse domain -> "techcrunch.com"
Step 2: ExchangeRegistry lookup
- Cache miss for techcrunch.com
- Background: fetch https://techcrunch.com/.well-known/fora.json
- - Discovers: exchange.ssp-alpha.com is authorized (PROVIDER_RELATIONSHIP_DIRECT)
- - Caches: techcrunch.com -> [ssp-alpha]
+ - Discovers: exchange.alpha.example is authorized (PROVIDER_RELATIONSHIP_DIRECT)
+ - Caches: techcrunch.com -> [exchange.alpha.example]
Step 3: Prepare RFC 9421 HTTP Message Signature (alg=ed25519)
- requester.id = "claude-agent-001"
@@ -251,13 +251,13 @@ Step 3: Prepare RFC 9421 HTTP Message Signature (alg=ed25519)
## Phase 2: Supply Discovery (Requester Identity, Subscription Detection, Content Attestations, Ed25519 Verification)
-The SDK sends `DiscoverResources` to SSP-Alpha — `POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/DiscoverResources`:
+The SDK sends `DiscoverResources` to Exchange Alpha — `POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/DiscoverResources`:
{/* fora-validate: ResourceQuery */}
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"requester": {
"id": "claude-agent-001",
"domain": "claude.ai",
@@ -270,7 +270,7 @@ The SDK sends `DiscoverResources` to SSP-Alpha — `POST https://exchange.ssp-al
}
```
-### Inside SSP-Alpha's Exchange
+### Inside Exchange Alpha
```
1. Validate request (proto validation) -> pass
@@ -341,11 +341,11 @@ The SDK sends `DiscoverResources` to SSP-Alpha — `POST https://exchange.ssp-al
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"offers": [
{
"offer_id": "2325749d-50d3-431a-ac34-900d443ee88e",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"title": "AI Regulation: What Providers Need to Know",
"pricing": {
"model": "PRICING_MODEL_PER_UNIT",
@@ -389,7 +389,7 @@ The SDK sends `DiscoverResources` to SSP-Alpha — `POST https://exchange.ssp-al
},
{
"offer_id": "4f155204-a53e-42aa-8471-154f87d7d3d5",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"title": "AI Regulation: What Providers Need to Know",
"pricing": {
"model": "PRICING_MODEL_FREE",
@@ -463,7 +463,7 @@ If Anthropic uses its hosted Broker instead of the SDK directly, the Broker prov
],
"acceptable_restrictions": [{ "axis": "RESTRICTION_KIND_FUNCTION", "values": ["ai-input"] }],
"constraints": {
- "preferred_exchanges": ["exchange.ssp-alpha.com"],
+ "preferred_exchanges": ["exchange.alpha.example"],
"max_unit_cost": "0.00005",
"reporting_capable": true,
"budget_scope": "team:claude-rag",
@@ -479,13 +479,13 @@ Discovery through the Broker is Broker-led: the agent asked the Broker, and the
1. Carries the agent's `requester` message, naming as `requester.domain` the host of the directory the agent's signature verified against (`claude.ai`, as the agent sent it)
2. Signs the query alone, with its own labeled RFC 9421 HTTP Message Signature covering its own request components and only its own `Signature-Agent` member. The query carries no agent signature headers
-Broker → SSP-Alpha (multi-URI batch query) — `POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/DiscoverResources`:
+Broker → Exchange Alpha (multi-URI batch query) — `POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/DiscoverResources`:
{/* fora-validate: ResourceQuery */}
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"requester": {
"id": "claude-agent-001",
"domain": "claude.ai",
@@ -501,21 +501,21 @@ Broker → SSP-Alpha (multi-URI batch query) — `POST https://exchange.ssp-alph
}
```
-SSP-Alpha's Exchange verifies the Broker's signature -- it looks up the public key by its `keyid` thumbprint in the key directory named by the Broker's `Signature-Agent` member, Ed25519 verifies, and authenticates the caller as the Broker. The query carries no delegation, and the Exchange would not honour one: nothing on a query the Broker signs alone can pass its holder binding. How an agent proves a holder-of-key entitlement at discovery time through a Broker is an open protocol question. A discovery query moves no money, so it carries no agent consent; see [Multi-hop forwarding](/protocol/authentication/#multi-hop-forwarding).
+Exchange Alpha verifies the Broker's signature -- it looks up the public key by its `keyid` thumbprint in the key directory named by the Broker's `Signature-Agent` member, Ed25519 verifies, and authenticates the caller as the Broker. The query carries no delegation, and the Exchange would not honour one: nothing on a query the Broker signs alone can pass its holder binding. How an agent proves a holder-of-key entitlement at discovery time through a Broker is an open protocol question. A discovery query moves no money, so it carries no agent consent; see [Multi-hop forwarding](/protocol/authentication/#multi-hop-forwarding).
-SSP-Alpha returns `offer_groups` (batch mode -- multiple URIs in query):
+Exchange Alpha returns `offer_groups` (batch mode -- multiple URIs in query):
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"offer_groups": [
{
"uri": "https://techcrunch.com/premium/ai-regulation-2026.html",
"offers": [
{
"offer_id": "2325749d-50d3-431a-ac34-900d443ee88e",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.05", "currency": "USD", "unit": "accesses", "unit_cost": "0.00001515", "estimated_quantity": 1 },
"identity": { "canonical_url": "https://techcrunch.com/premium/ai-regulation-2026.html", "iptc_guid": "urn:newsml:techcrunch:20260315:ai-reg-001", "content_hash": "a1b2c3d4e5f6...", "hash_method": "simhash-v1" },
"attestations": [{ "verifier": "gumgum.com", "claims": { "estimated_quantity": 3300, "word_count": 2500 }, "signature": "..." }],
@@ -529,7 +529,7 @@ SSP-Alpha returns `offer_groups` (batch mode -- multiple URIs in query):
},
{
"offer_id": "4f155204-a53e-42aa-8471-154f87d7d3d5",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"pricing": { "model": "PRICING_MODEL_FREE", "rate": "0", "currency": "USD", "unit_cost": "0", "estimated_quantity": 3300 },
"subscription_id": "SUB-ANTHROPIC-HEARST-2026",
"identity": { "canonical_url": "https://techcrunch.com/premium/ai-regulation-2026.html", "iptc_guid": "urn:newsml:techcrunch:20260315:ai-reg-001", "content_hash": "a1b2c3d4e5f6...", "hash_method": "simhash-v1" },
@@ -554,17 +554,17 @@ SSP-Alpha returns `offer_groups` (batch mode -- multiple URIs in query):
}
```
-Note: SSP-Alpha returns an empty `offers` array for the `theverge.com` URI with `absence_reason: NOT_IN_CATALOG` because The Verge is not an SSP-Alpha tenant. This v1.0 diagnostic lets the Broker distinguish "not in catalog" from "resource blocked for your use case." The Broker also queries SSP-Beta:
+Note: Exchange Alpha returns an empty `offers` array for the `theverge.com` URI with `absence_reason: NOT_IN_CATALOG` because The Verge is not an Exchange Alpha tenant. This v1.0 diagnostic lets the Broker distinguish "not in catalog" from "resource blocked for your use case." The Broker also queries Exchange Beta:
```json
-// SSP-Beta returns (single-URI, flat offers):
+// Exchange Beta returns (single-URI, flat offers):
{
"ver": "1.0",
- "exchange": "exchange.ssp-beta.com",
+ "exchange": "exchange.beta.example",
"offers": [
{
"offer_id": "ecb6fef4-1d83-4751-ab92-b3bc7c2fee39",
- "exchange": "exchange.ssp-beta.com",
+ "exchange": "exchange.beta.example",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.07", "currency": "USD", "unit": "accesses", "unit_cost": "0.00002258", "estimated_quantity": 1 },
"identity": {
"canonical_url": "https://theverge.com/premium/ai-regulation-republish.html",
@@ -595,8 +595,8 @@ Step 1: Group by iptc_guid
- These are ALL the same underlying article (TechCrunch original, syndicated to The Verge)
Step 2: SimHash cross-check (for non-news content without IPTC guids)
- - SSP-Alpha hash: "a1b2c3d4e5f6..." (simhash-v1)
- - SSP-Beta hash: "a1b2c3d4e5f7..." (simhash-v1)
+ - Exchange Alpha hash: "a1b2c3d4e5f6..." (simhash-v1)
+ - Exchange Beta hash: "a1b2c3d4e5f7..." (simhash-v1)
- Hamming distance = 1 bit -> substantially similar (threshold < 3 bits)
- For 99% of web content that lacks IPTC guids, SimHash is the primary dedup signal.
The Exchange computes SimHash during ingestion from extracted article text.
@@ -607,22 +607,15 @@ Step 2: SimHash cross-check (for non-news content without IPTC guids)
### Broker Selection Engine
```
-Step 3: Rank within content group
- Priority order:
- a) Subscription from preferred exchange
- -> 4f155204-a53e-42aa-8471-154f87d7d3d5 ($0, SSP-Alpha, attested by gumgum.com) <- WINNER
- b) Subscription from any exchange
- -> none
- c) Per-request from preferred exchange (rank by attestation trust, then unit_cost)
- -> 2325749d-50d3-431a-ac34-900d443ee88e ($0.05, SSP-Alpha, attested by gumgum.com)
- d) Per-request by unit_cost (all exchanges)
- -> ecb6fef4-1d83-4751-ab92-b3bc7c2fee39 ($0.07, SSP-Beta, attested by gumgum.com)
-
- Note: attestation claims enable selection beyond cheapest price.
- Offers with third-party attestations (Level 2) are preferred over
- self-attested (Level 1) or unattested (Level 0) content.
-
- Selected: 4f155204-a53e-42aa-8471-154f87d7d3d5 (subscription, zero marginal cost, third-party attested)
+Step 3: Order within the content group
+ The Broker keeps all three offers, each as the full package its Exchange
+ signed, and puts them in its order of relevance. The agent named Exchange Alpha
+ in preferred_exchanges, so Exchange Alpha's offers come first:
+ 1. 4f155204-a53e-42aa-8471-154f87d7d3d5 ($0, subscription, Exchange Alpha, attested by gumgum.com)
+ 2. 2325749d-50d3-431a-ac34-900d443ee88e ($0.05, per-request, Exchange Alpha, attested by gumgum.com)
+ 3. ecb6fef4-1d83-4751-ab92-b3bc7c2fee39 ($0.07, per-request, Exchange Beta, attested by gumgum.com)
+
+ Price is not part of the order. The agent compares the packages itself.
Step 4: Budget filter (applied before the offers are returned)
- Spend so far for scope "team:claude-rag", as this Broker counts it: $142.50
@@ -633,15 +626,19 @@ Step 4: Budget filter (applied before the offers are returned)
- Subscription offer: cost = $0 -> fits any budget
```
+### The Agent Chooses
+
+The agent gets all three offers and decides by its own policy: use its subscription where it covers the content, and prefer third-party attested content. It chooses `4f155204-a53e-42aa-8471-154f87d7d3d5`, the subscription offer from Exchange Alpha, at zero marginal cost and attested by gumgum.com.
+
---
## Phase 3: Transaction (Requester + Request Signature, Subscription Skip, Quota Check)
-SDK or Broker commits to the subscription offer.
+The agent commits to the subscription offer, directly with the SDK or through the Broker.
### Direct (SDK) request -- single signature in the headers:
-`POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/ExecuteTransaction`
+`POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/ExecuteTransaction`
{/* fora-validate: TransactionRequest */}
```json
@@ -658,7 +655,7 @@ SDK or Broker commits to the subscription offer.
{
"offer": {
"offer_id": "4f155204-a53e-42aa-8471-154f87d7d3d5",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"title": "AI Regulation: What Providers Need to Know",
"pricing": {
"model": "PRICING_MODEL_FREE",
@@ -696,9 +693,9 @@ SDK or Broker commits to the subscription offer.
### Orchestrated request -- bought through the Broker:
-The agent sends the purchase to the Broker, `POST https://broker.example/fora.v1.BrokerService/ExecuteTransaction`, signed for the Broker's URL. The Broker re-packages it and sends SSP-Alpha's Exchange its own sub-request:
+The agent sends the purchase to the Broker, `POST https://broker.example/fora.v1.BrokerService/ExecuteTransaction`, signed for the Broker's URL. The Broker re-packages it and sends Exchange Alpha its own sub-request:
-`POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/ExecuteTransaction`
+`POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/ExecuteTransaction`
{/* fora-validate: TransactionRequest */}
```json
@@ -715,7 +712,7 @@ The agent sends the purchase to the Broker, `POST https://broker.example/fora.v1
{
"offer": {
"offer_id": "4f155204-a53e-42aa-8471-154f87d7d3d5",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"title": "AI Regulation: What Providers Need to Know",
"pricing": {
"model": "PRICING_MODEL_FREE",
@@ -751,9 +748,9 @@ The agent sends the purchase to the Broker, `POST https://broker.example/fora.v1
}
```
-The sub-request body carries SSP-Alpha's items exactly as the agent signed them, with the agent's `agent_acceptance` on each item, its `agent_request_acceptance`, its `requester` and its `idempotency_key` unchanged. The Broker re-packaged this request for SSP-Alpha and signs it alone, with its own labeled RFC 9421 signature covering its own request components and only its own `Signature-Agent` member. The Exchange authenticates the caller as the Broker (its public key selected by `keyid` thumbprint in the directory the Broker's `Signature-Agent` member names). The agent's identity and consent are verified from the body: each `agent_acceptance` is a detached Ed25519 signature over `{offer_sig, requester_id, requester_domain, idempotency_key}`, which the Exchange verifies against the agent's registered key, resolved from `requester.domain` (`claude.ai/.well-known/http-message-signatures-directory`), and the request acceptance proves these items are the complete in-order set addressed to SSP-Alpha. That key also gives the delivery-URL binding, and a delegation's `cnf.jkt` would be checked against it. The Broker returns SSP-Alpha's answer to the agent inside a `BrokerTransactionResponse`, alongside the results of any other Exchange in the same purchase.
+The sub-request body carries Exchange Alpha's items exactly as the agent signed them, with the agent's `agent_acceptance` on each item, its `agent_request_acceptance`, its `requester` and its `idempotency_key` unchanged. The Broker re-packaged this request for Exchange Alpha and signs it alone, with its own labeled RFC 9421 signature covering its own request components and only its own `Signature-Agent` member. The Exchange authenticates the caller as the Broker (its public key selected by `keyid` thumbprint in the directory the Broker's `Signature-Agent` member names). The agent's identity and consent are verified from the body: each `agent_acceptance` is a detached Ed25519 signature over `{offer_sig, requester_id, requester_domain, idempotency_key}`, which the Exchange verifies against the agent's registered key, resolved from `requester.domain` (`claude.ai/.well-known/http-message-signatures-directory`), and the request acceptance proves these items are the complete in-order set addressed to Exchange Alpha. That key also gives the delivery-URL binding, and a delegation's `cnf.jkt` would be checked against it. The Broker returns Exchange Alpha's answer to the agent inside a `BrokerTransactionResponse`, alongside the results of any other Exchange in the same purchase.
-### Inside SSP-Alpha's Exchange
+### Inside Exchange Alpha
```
1. Validate request (proto validation) -> pass
@@ -888,13 +885,13 @@ CDN access log records:
The SDK auto-submits a usage report (background, non-blocking). Usage reporting is mandatory for subscription transactions -- failure to report within the window triggers `DENIAL_REASON_REPORTING_OVERDUE` on subsequent transactions.
-`POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/ReportUsage`
+`POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/ReportUsage`
{/* fora-validate: UsageReport */}
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"idempotency_key": "ur-claude-001",
"transaction_id": "txn-alpha-001",
"billing_id": "bill-sub-alpha-001",
@@ -952,7 +949,7 @@ Three independent records that must agree:
- url_hash=9f2c7a4e... fetched at 15:25:00, 200 OK, 45KB
- agent_id=NzbLsXh8..., signature and agent binding verified by the edge
-2. Exchange transaction log (SSP-Alpha controls):
+2. Exchange transaction log (Exchange Alpha controls):
- txn-alpha-001: signed_url_hash=9f2c7a4e... (the join key to the log above)
- subscription SUB-ANTHROPIC-HEARST-2026, amount=$0
- subscription_unit_value: $0.05 (value of the access for accounting)
@@ -981,7 +978,7 @@ PASS: Citation included as required by the Offer's attribution terms
Hearst queries the Exchange's provider audit API to independently verify all transactions for their content:
```
-GET https://exchange.ssp-alpha.com/provider/techcrunch.com/transactions
+GET https://exchange.alpha.example/provider/techcrunch.com/transactions
?from=2026-03-15T00:00:00Z
&to=2026-03-16T00:00:00Z
```
@@ -996,7 +993,7 @@ Response includes signed Offer snapshots for every transaction:
"billing_id": "bill-sub-alpha-001",
"offer_snapshot": {
"offer_id": "4f155204-a53e-42aa-8471-154f87d7d3d5",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"pricing": { "model": "PRICING_MODEL_FREE", "rate": "0", "currency": "USD" },
"subscription_id": "SUB-ANTHROPIC-HEARST-2026",
"terms": [
@@ -1018,11 +1015,11 @@ Response includes signed Offer snapshots for every transaction:
**What Hearst can verify**:
-1. **Offer authenticity**: Hearst verifies the `exchange_signature` on each `offer_snapshot` using SSP-Alpha's published Ed25519 public key (fetch SSP-Alpha's WBA directory at `exchange.ssp-alpha.com/.well-known/http-message-signatures-directory` and read its `keys`). This proves the Exchange actually issued this Offer -- it cannot deny having offered these terms.
+1. **Offer authenticity**: Hearst verifies the `exchange_signature` on each `offer_snapshot` using Exchange Alpha's published Ed25519 public key (fetch Exchange Alpha's WBA directory at `exchange.alpha.example/.well-known/http-message-signatures-directory` and read its `keys`). This proves the Exchange actually issued this Offer -- it cannot deny having offered these terms.
2. **RSL price ceiling compliance**: Hearst's RSL declares a maximum rate of $0.08/crawl. The subscription's `subscription_unit_value` of $0.05 is below this ceiling. If any per-request Offer exceeded $0.08, Hearst would detect it by comparing `offer_snapshot.pricing.rate` against their RSL terms.
-3. **Private floor compliance**: Hearst's contract with SSP-Alpha specifies a private minimum rate of $0.04/article. The `subscription_unit_value` of $0.05 is above this floor. Private floor pricing is contractual (not protocol-enforced), but the audit data makes violations detectable.
+3. **Private floor compliance**: Hearst's contract with Exchange Alpha specifies a private minimum rate of $0.04/article. The `subscription_unit_value` of $0.05 is above this floor. Private floor pricing is contractual (not protocol-enforced), but the audit data makes violations detectable.
### What Hearst Sees (Subscription Summary)
@@ -1062,7 +1059,7 @@ But if Hamming distance > threshold, or content is entirely different:
### Dispute Request
```json
-POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/DisputeTransaction
+POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/DisputeTransaction
{
"ver": "1.0",
@@ -1093,14 +1090,14 @@ Three-tier resolution: Tier 1 automated (<1s, auto-disputable claims like CDN
## What the Broker Adds (Cross-Provider Intelligence)
-**Without Broker (SDK only)**: the agent found the article via SSP-Alpha (from `fora.json`) and got the subscription deal. It never knew The Verge also had the article at $0.07 via SSP-Beta.
+**Without Broker (SDK only)**: the agent found the article via Exchange Alpha (from `fora.json`) and got the subscription deal. It never knew The Verge also had the article at $0.07 via Exchange Beta.
-**With Broker**: it queried both Exchanges, discovered the same article via iptc_guid match (confirmed by SimHash similarity), received attestation data from both sources, and still chose the subscription offer -- but now Anthropic has intelligence:
+**With Broker**: the Broker queried both Exchanges, matched the same article via iptc_guid (confirmed by SimHash similarity) and returned every offer with its attestation data. The agent still chose the subscription offer -- but now Anthropic has intelligence:
```
"This article is available from two sources:
- - SSP-Alpha (TechCrunch, subscription, $0/request, third-party attested by gumgum.com)
- - SSP-Beta (The Verge, per-request, $0.07/request, third-party attested by gumgum.com)
+ - Exchange Alpha (TechCrunch, subscription, $0/request, third-party attested by gumgum.com)
+ - Exchange Beta (The Verge, per-request, $0.07/request, third-party attested by gumgum.com)
Our subscription saves us $0.07 per request from the alternative source.
At 10K requests/month for this provider's content, that's $700/month
@@ -1110,4 +1107,4 @@ Both sources carry Level 2 (third-party) attestations from gumgum.com.
TechCrunch original has higher word_count (2500 vs 2350) — more complete."
```
-This is the Broker's value: **not just cheapest price, but market intelligence for subscription negotiations and attestation-aware content selection.**
+This is the Broker's value: **every offer for the same content in one answer, each as its Exchange signed it, so the agent decides with the whole picture: market intelligence for subscription negotiations and attestation-aware content selection.**
diff --git a/website/src/content/docs/protocol/standards-layering.mdx b/website/src/content/docs/protocol/standards-layering.mdx
index c83ff8bc..58acbfc4 100644
--- a/website/src/content/docs/protocol/standards-layering.mdx
+++ b/website/src/content/docs/protocol/standards-layering.mdx
@@ -93,7 +93,7 @@ graph LR
| Role | Description | Standard |
|---|---|---|
| **AI Agent** | The LLM, RAG system, or agentic application that needs content. Authenticates via Ed25519 signatures. Identity is the RFC 9421 request signature; the scopes it states decide which terms it is shown, and a delegation can back them where an Exchange requires one. | CoMP: "Requesting Party". FORA v1.0 uses the `Requester` message (not CoMP `AISystem`); CoMP's `lid` is a public agent identifier with no counterpart field on the FORA request — not a credential. |
-| **FORA Broker** | Optional demand-side router. Queries multiple Exchanges, compares offers by unit_cost, deduplicates by ResourceIdentity, enforces budget constraints. | FORA v1.0: "FORA Broker" |
+| **FORA Broker** | Optional demand-side router. Queries multiple Exchanges, groups offers by ResourceIdentity, returns them in its order of relevance, applies the agent's budget constraints. | FORA v1.0: "FORA Broker" |
| **Exchange** | Exposes resource catalog with pricing. Authorizes transactions. Issues delivery instructions (signed URLs). | FORA v1.0: "Exchange"; implementation of CoMP endpoints |
| **Resource Provider** | The provider. Owns the resources, sets terms via RSL. | CoMP: "Content Owner"; RSL: "Licensor" |
| **Resource Delivery Endpoint** | The CDN that serves resources. Verifies signed URLs natively. | CoMP: "Content Delivery Endpoint" |
@@ -300,8 +300,8 @@ An Exchange serves its `fora.json` (`role=ROLE_EXCHANGE`) declaring:
],
"exchanges": [
{
- "domain": "exchange.ssp-example.com",
- "endpoint": "https://exchange.ssp-example.com/fora/v1",
+ "domain": "exchange.example",
+ "endpoint": "https://exchange.example/fora/v1",
"relationship": "DIRECT"
}
]
diff --git a/website/src/content/docs/protocol/transaction-flow.mdx b/website/src/content/docs/protocol/transaction-flow.mdx
index 0fd59697..f1403da4 100644
--- a/website/src/content/docs/protocol/transaction-flow.mdx
+++ b/website/src/content/docs/protocol/transaction-flow.mdx
@@ -9,7 +9,7 @@ The Exchange exposes four RPCs. This is the core protocol -- everything else is
```
With Broker (multi-exchange, budget control, reporting):
- Agent -> Broker -> Exchange A, B, C -> select best
+ Agent -> Broker -> Exchange A, B, C -> offers, in order of relevance
Agent -> Exchange A, B, C directly -> buy, deliver
or Agent -> Broker -> Exchange A, B, C -> buy in one call, deliver
@@ -44,7 +44,7 @@ Content-Type: application/json
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"requester": {
"id": "research-bot",
"domain": "buyer.example.com",
@@ -70,11 +70,11 @@ Key fields:
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"offers": [
{
"offer_id": "9d5735d7-1f17-4801-8dcf-7cac0b7b32b0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"title": "The Future of AI Infrastructure",
"ext": {
"comp.package_id": "PKG-TC-AI-INFRA",
@@ -147,14 +147,14 @@ When requesting multiple URIs, the response uses `offer_groups` instead of `offe
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"offer_groups": [
{
"uri": "https://cdn.fora-protocol.org/premium/ai-infrastructure.html",
"offers": [
{
"offer_id": "9d5735d7-1f17-4801-8dcf-7cac0b7b32b0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.00002", "currency": "USD", "unit_cost": "0.00002", "estimated_quantity": 2500, "unit": "tokens" },
"terms": [
{
@@ -171,7 +171,7 @@ When requesting multiple URIs, the response uses `offer_groups` instead of `offe
"offers": [
{
"offer_id": "89636cda-93f6-460e-af7f-2b6aef0f491b",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.00002", "currency": "USD", "unit_cost": "0.00002", "estimated_quantity": 2500, "unit": "tokens" },
"terms": [
{
@@ -202,7 +202,7 @@ When the requester has a subscription, the Exchange includes both per-request an
"offers": [
{
"offer_id": "9d5735d7-1f17-4801-8dcf-7cac0b7b32b0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.00002", "currency": "USD", "unit_cost": "0.00002", "estimated_quantity": 2500, "unit": "tokens" },
"terms": [
{
@@ -214,7 +214,7 @@ When the requester has a subscription, the Exchange includes both per-request an
},
{
"offer_id": "c5b01a4e-6536-4a42-a4fc-e41a1df1913d",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"pricing": { "model": "PRICING_MODEL_FREE", "rate": "0", "currency": "USD", "unit_cost": "0" },
"subscription_id": "SUB-12345",
"terms": [ { "semantics": "TERM_SEMANTICS_ENUMERATED", "scopes": ["subscription:SUB-12345"] } ],
@@ -253,7 +253,7 @@ When the requester holds a subscription, the Exchange attaches `SubscriptionQuot
"offers": [
{
"offer_id": "c5b01a4e-6536-4a42-a4fc-e41a1df1913d",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"pricing": { "model": "PRICING_MODEL_FREE", "rate": "0", "currency": "USD" },
"subscription_id": "SUB-12345",
"terms": [ { "semantics": "TERM_SEMANTICS_ENUMERATED", "scopes": ["subscription:SUB-12345"] } ],
@@ -310,7 +310,7 @@ Content-Type: application/json
{
"offer": {
"offer_id": "9d5735d7-1f17-4801-8dcf-7cac0b7b32b0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"title": "The Future of AI Infrastructure",
"pricing": {
"model": "PRICING_MODEL_PER_UNIT",
@@ -436,7 +436,7 @@ For batch transactions, use the `items` array:
{
"offer": {
"offer_id": "9d5735d7-1f17-4801-8dcf-7cac0b7b32b0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.00002", "currency": "USD", "unit_cost": "0.00002", "estimated_quantity": 2500, "unit": "tokens" },
"expires_at": "2026-03-14T02:30:00Z",
"terms": [
@@ -451,7 +451,7 @@ For batch transactions, use the `items` array:
{
"offer": {
"offer_id": "89636cda-93f6-460e-af7f-2b6aef0f491b",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.00002", "currency": "USD", "unit_cost": "0.00002", "estimated_quantity": 2500, "unit": "tokens" },
"expires_at": "2026-03-14T02:30:00Z",
"terms": [
@@ -607,7 +607,7 @@ Content-Type: application/json
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"idempotency_key": "ur-e4f2a1",
"transaction_id": "txn-mp-93a7f2",
"billing_id": "bill-93a7f2-001",
@@ -717,11 +717,11 @@ sequenceDiagram
PUB->>MP: Compare: signed URL accesses vs transactions
```
-Three independent records -- CDN logs, Exchange transactions, usage reports -- that must agree. This provides stronger reconciliation than ad-tech's two-sided model.
+Three independent records -- CDN logs, Exchange transactions, usage reports -- that must agree.
## Consistency Model
-FORA uses **at-most-once with reconciliation**, same as OpenRTB:
+FORA uses **at-most-once with reconciliation**:
- Idempotency keys on every `TransactionRequest`, scoped to the agent whatever the route (prevent double-charge on retry, through a Broker or directly; a key reused with different items is refused with `already_exists`)
- "Authorize first, deliver second" ordering (do not issue signed URL until billing succeeds)
@@ -743,7 +743,7 @@ Content-Type: application/json
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"idempotency_key": "dsp-20260318-abc123",
"transaction_id": "txn-mp-93a7f2",
"billing_id": "bill-93a7f2-001",
@@ -811,7 +811,7 @@ message Preview {
```json
{
"offer_id": "9bb01145-326e-42a9-b2ac-058e3114bdfc",
- "exchange": "exchange.ssp-example.com",
+ "exchange": "exchange.example",
"title": "Climate Summit Photo - Reuters",
"pricing": { "model": "PRICING_MODEL_PER_UNIT", "rate": "0.50", "currency": "USD", "estimated_quantity": 1, "unit": "accesses" },
"previews": [
@@ -838,7 +838,6 @@ This follows the universal content exchange pattern:
- **Shutterstock**: `small_thumb.url` (100px, unwatermarked), `preview.url` (450px, watermarked), `preview_1500.url` (1500px, watermarked)
- **Spotify**: `preview_url` -- URL to a 30-second MP3 clip
- **IIIF**: Parameterized URL template -- client requests any size
-- **OpenRTB**: `img.url` + `img.w` + `img.h` per asset
Nobody puts image bytes in offer metadata. The Exchange holds a URL string; the CDN holds the pixels.
diff --git a/website/src/content/docs/protocol/walkthrough-academic.mdx b/website/src/content/docs/protocol/walkthrough-academic.mdx
index 59358ac6..6a831ac2 100644
--- a/website/src/content/docs/protocol/walkthrough-academic.mdx
+++ b/website/src/content/docs/protocol/walkthrough-academic.mdx
@@ -22,7 +22,7 @@ The agent carries a delegation JWT from the university, granting `academic:*` an
**Seven-step flow:**
```
-Discovery → ResourceQuery (Broker-led) → ResourceResponse → Direct ResourceQuery with the delegation → Broker Selection → Batch ExecuteTransaction → ReportUsage
+Discovery → ResourceQuery (Broker-led) → ResourceResponse → Direct ResourceQuery with the delegation → Agent Selection → Batch ExecuteTransaction → ReportUsage
```
---
@@ -542,8 +542,8 @@ The Exchange classified each of the 50 DOIs into one of three categories based o
Because the ResourceQuery declared `supported_profiles: ["fora-academic-v1"]`, every offer includes academic-specific metadata:
-- **`academic.oa_status`** — Open Access classification. Values: `gold` (fully OA), `green` (author self-archived), `hybrid` (OA in subscription journal), `bronze` (free-to-read but no license), `diamond` (no APC), `closed` (paywalled). The Broker uses this to prefer free sources.
-- **`academic.version`** — Article lifecycle stage. Values: `preprint` (author's original), `aam` (accepted author manuscript), `vor` (Version of Record, publisher-formatted). The Broker prefers VoR when available, falls back to preprint.
+- **`academic.oa_status`** — Open Access classification. Values: `gold` (fully OA), `green` (author self-archived), `hybrid` (OA in subscription journal), `bronze` (free-to-read but no license), `diamond` (no APC), `closed` (paywalled). The agent can use this to prefer free sources.
+- **`academic.version`** — Article lifecycle stage. Values: `preprint` (author's original), `aam` (accepted author manuscript), `vor` (Version of Record, publisher-formatted). The Broker puts VoR first when available, then the preprint.
- **`academic.retraction_status`** — Publication integrity status. Values: `active` (valid), `retracted` (withdrawn), `expression_of_concern` (under investigation). When `retracted` appears in `ext_critical`, the consumer MUST understand retraction semantics or reject the offer.
- **`academic.peer_review_status`** — Whether the paper has been peer-reviewed. Values: `peer_reviewed`, `not_peer_reviewed`. Useful for weighting evidence quality.
@@ -559,7 +559,7 @@ The retracted paper (DOI `10.1038/s41592-024-02467-x`) is immediately identifiab
### Content Identity and Dedup
Every offer carries a `ResourceIdentity` with:
-- **`doi`** — The DOI serves as the universal dedup key. If the same paper were available from multiple Exchanges (e.g., via a different aggregator), the Broker groups offers by DOI to compare pricing.
+- **`doi`** — The DOI serves as the universal dedup key. If the same paper were available from multiple Exchanges (e.g., via a different aggregator), the Broker groups offers by DOI, so the agent sees every offer for the paper together.
- **`content_hash`** — SHA-256 hash for post-delivery verification. Because `resource_mutability` is `RESOURCE_MUTABILITY_STATIC` for all academic papers (published articles do not change — corrections and retractions create new resources), the agent SHOULD verify the hash on delivery.
- **`canonical_url`** — The publisher's authoritative URL. Different from the DOI URI but resolves to the same resource.
@@ -747,16 +747,16 @@ For the 20 Elsevier papers covered by Stanford's institutional subscription:
---
-## Step 5 — Broker Selection
+## Step 5 — Agent Selection
-The agent hands the direct answer from Step 4 to the Broker, which applies a cost-minimization strategy across all 50 OfferGroups, with the 20 subscription offers in place of the paid Elsevier offers for those DOIs:
+The agent chooses the offers it buys by its own policy, set by the research lab: spend as little as the review allows. It applies that policy across all 50 OfferGroups, with the 20 subscription offers from Step 4 in place of the paid Elsevier offers for those DOIs. The Broker returned the offers in its order of relevance; the choice below is the agent's:
### Selection Rules
1. **Exclude retracted papers** — The retracted paper is flagged to the research agent. The agent decides whether to exclude it from the review or include it with a retraction annotation.
2. **Prefer free OA** — 12 arXiv papers at $0.00 each.
3. **Prefer subscription** — 20 Elsevier papers at $0.00 each (the subscription offers from Step 4).
-4. **Cheapest paid for the rest** — 17 papers at $1.50 each from Springer and Elsevier (no subscription coverage).
+4. **Lowest price for the rest, by the lab's policy** — 17 papers at $1.50 each from Springer and Elsevier (no subscription coverage).
### Cost Summary
@@ -768,11 +768,11 @@ The agent hands the direct answer from Step 4 to the Broker, which applies a cos
| Retracted (excluded) | 1 | — | $0.00 |
| **Total** | **49 acquired** | | **$25.50** |
-Without the institutional subscription, all 50 papers at $1.50 each would cost $75.00. The institutional subscription saved $30.00 (40%), and free OA saved another $18.00 (24%). The Broker reports the savings breakdown to the research agent.
+Without the institutional subscription, all 50 papers at $1.50 each would cost $75.00. The institutional subscription saved $30.00 (40%), and free OA saved another $18.00 (24%).
-The Broker also flags the retracted paper to the agent:
+The Broker flags the retracted paper to the agent:
-> **Warning:** DOI `10.1038/s41592-024-02467-x` ("Attention-Guided Pathology Detection via Self-Supervised Transformers") was retracted on 2026-02-15. Reason: data fabrication. This paper has been excluded from the batch transaction. Include it explicitly if needed for retraction analysis.
+> **Warning:** DOI `10.1038/s41592-024-02467-x` ("Attention-Guided Pathology Detection via Self-Supervised Transformers") was retracted on 2026-02-15. Reason: data fabrication. Leave it out of the batch transaction, or include it explicitly if needed for retraction analysis.
---
@@ -1031,7 +1031,7 @@ Response:
| 2 | `DiscoverResources` (Broker-led) | `Requester` stating `academic:*` | Subscription term hidden: the delegation is not honored on the Broker's query | Same | Same |
| 3 | -- | Offers at $0, `oa_status: "gold"` | Paid offers at $1.50 only | Offers at $1.50 | Offer with `retraction_status: "retracted"` in `ext_critical` |
| 4 | `DiscoverResources` (direct, with the delegation) | -- | Delegation verified; offers at $0, `subscription_id` set | Unchanged, $1.50 | -- |
-| 5 | -- | Broker selects (free) | Broker selects (free via sub) | Broker selects (cheapest) | Broker flags and excludes |
+| 5 | -- | Agent selects (free) | Agent selects (free via sub) | Agent selects (its cost policy) | Broker flags; agent excludes |
| 6 | `ExecuteTransaction` | Batch, cost $0 | Batch, cost $0, sub quota decremented | Batch, cost $1.50/ea | Not transacted |
| 7 | `ReportUsage` | One report per transaction, in tokens | One report per transaction, in tokens | One report per transaction, 1 access each | No report |
diff --git a/website/src/content/docs/protocol/walkthrough-v1.mdx b/website/src/content/docs/protocol/walkthrough-v1.mdx
index cacd049a..96c23ed8 100644
--- a/website/src/content/docs/protocol/walkthrough-v1.mdx
+++ b/website/src/content/docs/protocol/walkthrough-v1.mdx
@@ -44,8 +44,8 @@ curl -s https://techcrunch.com/.well-known/fora.json | jq
"contact": "licensing@techcrunch.com",
"exchanges": [
{
- "domain": "exchange.ssp-alpha.com",
- "endpoint": "https://exchange.ssp-alpha.com/fora/v1",
+ "domain": "exchange.alpha.example",
+ "endpoint": "https://exchange.alpha.example/fora/v1",
"relationship": "PROVIDER_RELATIONSHIP_DIRECT"
}
],
@@ -59,13 +59,13 @@ The agent now knows which Exchange to query for TechCrunch content. The `support
The agent sends a `ResourceQuery` to the Exchange with its identity in the `Requester` message. This is the v1.0 replacement for CoMP's `AISystem` — it carries identity, scopes, and optional delegation in a single envelope. The request is authenticated separately by an RFC 9421 HTTP Message Signature in the HTTP headers, not by a field in the message body.
-`POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/DiscoverResources`
+`POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/DiscoverResources`
{/* fora-validate: ResourceQuery */}
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"uris": [
"https://techcrunch.com/2026/03/19/ai-agents-commerce.html"
],
@@ -97,14 +97,14 @@ The Exchange looks up the article in its catalog and returns one offer for each
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"offer_groups": [
{
"uri": "https://techcrunch.com/2026/03/19/ai-agents-commerce.html",
"offers": [
{
"offer_id": "3a1cf96e-74b7-4c5d-abb0-925c3f83dfe3",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"title": "AI Agents Are Rewriting Commerce",
"ext": {
"comp.package_id": "PKG-TC-AGENTS-COMMERCE",
@@ -179,7 +179,7 @@ The Exchange looks up the article in its catalog and returns one offer for each
The agent commits to the offer by sending the full signed `offer` back, reflected exactly as received at discovery:
-`POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/ExecuteTransaction`
+`POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/ExecuteTransaction`
{/* fora-validate: TransactionRequest */}
```json
@@ -190,7 +190,7 @@ The agent commits to the offer by sending the full signed `offer` back, reflecte
{
"offer": {
"offer_id": "3a1cf96e-74b7-4c5d-abb0-925c3f83dfe3",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"title": "AI Agents Are Rewriting Commerce",
"pricing": {
"model": "PRICING_MODEL_PER_UNIT",
@@ -288,13 +288,13 @@ The CDN edge function verifies:
Usage reporting is mandatory. The agent reports actual consumption, in the unit of the offer's price (`accesses`):
-`POST https://exchange.ssp-alpha.com/fora/v1/fora.v1.ExchangeService/ReportUsage`
+`POST https://exchange.alpha.example/fora/v1/fora.v1.ExchangeService/ReportUsage`
{/* fora-validate: UsageReport */}
```json
{
"ver": "1.0",
- "exchange": "exchange.ssp-alpha.com",
+ "exchange": "exchange.alpha.example",
"idempotency_key": "ur-001",
"transaction_id": "txn-tc-001",
"billing_id": "bill-tc-001",
diff --git a/website/src/content/docs/reference/changelog.mdx b/website/src/content/docs/reference/changelog.mdx
index 7dc5b581..9f452ebe 100644
--- a/website/src/content/docs/reference/changelog.mdx
+++ b/website/src/content/docs/reference/changelog.mdx
@@ -8,6 +8,36 @@ and protocol history, see [`proto/CHANGELOG.md`](https://github.com/FORA-Protoco
## Unreleased
+**A Broker orders offers by relevance, and the agent decides (docs and comments only; no wire
+change).** The docs no longer describe the Broker as choosing the cheapest offer or the best deal.
+A Broker returns every Exchange's signed offers in its order of relevance to the agent's request,
+optionally narrowed to the kinds of license the agent accepts (`acceptable_restrictions`). It
+groups offers for the same content without dropping any, and it never buys, resells or puts one
+offer in place of another. Price is not a signal: each offer carries its price and
+`Pricing.unit_cost`, and the agent compares them under its own policy. The Selection Engine page
+is rewritten on these lines, and the Content Deduplication, Discovery Paths, walkthrough, Agent
+SDK, money-flow and operator pages follow. Two comments in `fora.proto` (`Offer.pricing` and the
+one-term-per-offer note) now say an agent compares the price, not that a Broker ranks on it.
+
+**The edge function refuses AI search crawlers by default (docs only; no wire change).** Bot
+Detection now describes the lists FORA Server ships: AI search crawlers such as PerplexityBot
+get a `403` on licensed paths by default, and a provider changes the built-in lists with
+`BOT_UA_ALLOW_JSON` and `BOT_UA_DENY_JSON`. The `bot_detection.ai_search_policy` setting the page
+described doesn't exist and is gone. Googlebot, Bingbot and Applebot are never refused. The
+publishers page, the AI crawlers resource, the bot table and the rules generator say the same.
+
+**Publisher onboarding names the console's operator (docs only).** FORA is open-source software,
+not an operator. The publisher console on fora-protocol.org is run by Postindustria, which
+charges publishers no Exchange fee beyond the payment system's transaction fees.
+
+**Proto comments state the SSRF guards instead of citing the threat model (comments only; no
+wire change).** The `License.uri` comment and the CoMP comments on `Package.licenseurl`,
+`Package.reporturl` and `Retrieval.endpoint` cited threat ID T-LIC-1 in the threat model. They
+now state the guards themselves: a scheme allowlist, blocking loopback, private and metadata
+addresses (resolve-then-check), fetching through an egress proxy, and treating the response as
+untrusted content. The CoMP comments say these are the guards specified for `License.uri`. The
+`DiscoveryResponse.absence_reason` comment no longer ends "See the threat model."
+
**The fora-music-v1 search filter keys are a published JSON Schema (docs and schema; no wire
change).** The 23 keys of `DiscoveryRequest.search_filters` are one file,
`schemas/music/v1/music-v1.search-filters.schema.json`, published at
diff --git a/website/src/content/docs/reference/fora-json-example.mdx b/website/src/content/docs/reference/fora-json-example.mdx
index d691dbc5..44343a81 100644
--- a/website/src/content/docs/reference/fora-json-example.mdx
+++ b/website/src/content/docs/reference/fora-json-example.mdx
@@ -25,13 +25,13 @@ The `fora.json` file is served at `/.well-known/fora.json` and carries a `WellKn
],
"exchanges": [
{
- "domain": "exchange.ssp-alpha.com",
- "endpoint": "https://exchange.ssp-alpha.com/v1",
+ "domain": "exchange.alpha.example",
+ "endpoint": "https://exchange.alpha.example/v1",
"relationship": "DIRECT"
},
{
- "domain": "exchange.ssp-beta.com",
- "endpoint": "https://exchange.ssp-beta.com/v1",
+ "domain": "exchange.beta.example",
+ "endpoint": "https://exchange.beta.example/v1",
"relationship": "RESELLER"
}
]
@@ -99,7 +99,7 @@ The agent checks `/.well-known/fora.json` before attempting to access content:
```
1. Agent wants content from techprovider.com
2. GET https://techprovider.com/.well-known/fora.json
-3. Finds Exchange endpoint: exchange.ssp-alpha.com/v1
+3. Finds Exchange endpoint: exchange.alpha.example/v1
4. Calls DiscoverResources → gets Offers with pricing
5. Calls ExecuteTransaction → gets signed URL
6. Fetches content from CDN
@@ -113,7 +113,7 @@ If the agent doesn't know about FORA and tries to crawl directly:
1. Agent hits techprovider.com/premium/article
2. Edge function returns 403 with two headers:
X-Content-Rules: https://techprovider.com/.well-known/fora.json
- X-FORA-Exchange: exchange.ssp-alpha.com
+ X-FORA-Exchange: exchange.alpha.example
3. Agent discovers fora.json from X-Content-Rules (and, optionally, starts
discovery at the X-FORA-Exchange domain, resolving its endpoint from
that Exchange's own fora.json)
@@ -123,10 +123,10 @@ If the agent doesn't know about FORA and tries to crawl directly:
## Relationship Types
-| Relationship | Description | Ad-Tech Equivalent |
-|---|---|---|
-| `DIRECT` | Provider has a direct contract with this Exchange | ads.txt DIRECT |
-| `RESELLER` | Exchange resells content via another authorized party | ads.txt RESELLER |
+| Relationship | Description |
+|---|---|
+| `DIRECT` | Provider has a direct contract with this Exchange |
+| `RESELLER` | Exchange resells content via another authorized party |
## Provider Trust Model
diff --git a/website/src/content/docs/reference/proto-fora.mdx b/website/src/content/docs/reference/proto-fora.mdx
index 40e8a5de..8d9ad251 100644
--- a/website/src/content/docs/reference/proto-fora.mdx
+++ b/website/src/content/docs/reference/proto-fora.mdx
@@ -113,7 +113,7 @@ There is no in-message hop chain. Request authentication is carried in **RFC 942
::proto-message{name=Offer}
:::note[An offer carries its price once]
-`Offer.pricing` is the offer's only price: the price execute charges, a Broker ranks, and a metered purchase is charged at. The one term the offer sells, in `Offer.terms`, carries no `pricing` (rule `offer.terms.pricing_unset`). Catalog terms (`ResourceEntry.terms`) still carry their pricing; each offer derives from exactly one of them, so the offer's price and its term's price cannot disagree.
+`Offer.pricing` is the offer's only price: the price execute charges, an agent compares, and a metered purchase is charged at. The one term the offer sells, in `Offer.terms`, carries no `pricing` (rule `offer.terms.pricing_unset`). Catalog terms (`ResourceEntry.terms`) still carry their pricing; each offer derives from exactly one of them, so the offer's price and its term's price cannot disagree.
:::
:::note[A metered offer may carry an estimate]
diff --git a/website/src/data/collective-evidence.ts b/website/src/data/collective-evidence.ts
index e1b336df..78b7bf55 100644
--- a/website/src/data/collective-evidence.ts
+++ b/website/src/data/collective-evidence.ts
@@ -188,7 +188,7 @@ const Q = {
koPretraining: `This agreement is limited to regulating the National Library’s training of language models on the data governed by this agreement. It does not affect any pre-training that the base models used by the National Library may have undergone.`,
offerSig: `REQUIRED. Hex-encoded detached Ed25519 signature over the canonical serialization of the ENTIRE Offer — every field, including \`pricing\`, \`terms\` (the full licensing payload), \`expires_at\`, and \`exchange\`.`,
acceptance: `The agent's detached acceptance signature over this item's \`offer\`.`,
- money: `FORA never touches money. It follows the OpenRTB model: the protocol handles authorization and logging, while actual settlement happens outside the protocol through existing billing infrastructure.`,
+ money: `FORA never touches money. The protocol handles authorization and logging, while actual settlement happens outside the protocol through existing billing infrastructure.`,
manyTerms: `A resource carries zero or more \`LicenseTerm\` entries, at most 32. Each term is a complete, self-contained access arrangement, and a resource with several terms is the normal case.`,
brokerSearch: `A Broker runs one catalog search, which its operator chooses per deployment. The search only proposes URIs: prices, license terms and signatures always come from the Exchange, so a search result cannot produce an offer the Exchange does not stand behind.`,
ndpLicensed: `Members of NDP Nieuwsmedia are providing access to their archives, which include articles from over 30 national and regional news titles.`,
diff --git a/website/src/data/content-marketplaces.ts b/website/src/data/content-marketplaces.ts
new file mode 100644
index 00000000..d1261ec5
--- /dev/null
+++ b/website/src/data/content-marketplaces.ts
@@ -0,0 +1,262 @@
+// The content-marketplaces series (/resources/content-marketplaces/ and the
+// pages under it, plus /resources/standards-map/): the texts every page uses
+// word for word, so the seven pages say the same thing the same way. A page
+// imports a text from here and never types it again. Change a text here and
+// every page changes with it.
+//
+// Texts that hold links are HTML strings; the components render them with
+// set:html. "See About FORA." links to the About FORA card, which every page
+// carries with id="about".
+
+/** How real a thing is. StatusTag (src/lib/resource-markup.ts) renders these. */
+export const STATUS = {
+ live: { label: 'Live', means: 'Released and in use.' },
+ beta: { label: 'Beta', means: 'Open to sign up, labelled beta by the vendor.' },
+ 'closed-beta': { label: 'Closed beta', means: 'By admission only.' },
+ pilot: { label: 'Pilot', means: 'The vendor\'s own word for a limited program, such as Microsoft\'s PCM.' },
+ design: { label: 'Design', means: 'Designed, not in released software. Every FORA design item carries it.' },
+ roadmap: { label: 'Roadmap', means: 'Planned, not designed in public.' },
+ experimental: { label: 'Experimental', means: 'The vendor\'s own word for a feature it is testing, such as Google\'s Web Bot Auth signing.' },
+ preview: { label: 'Preview', means: 'The vendor\'s own word for a released feature it hasn\'t made generally available, such as AWS\'s.' },
+ 'not-public': { label: 'Not public', means: 'The vendor doesn\'t say.' },
+} as const;
+export type StatusKey = keyof typeof STATUS;
+
+/**
+ * The positioning statement and the tagline the owner approved on 2026-11-15
+ * (messaging review, option A), word for word. The About page and the
+ * content-marketplaces page both use them.
+ */
+export const POSITIONING = 'FORA lets any AI agent find your content, read your license, pay and get a signed receipt, on the terms you set. It is an open protocol and free, open-source software, so any company can also run its own FORA marketplace.';
+/** The tagline; the phrase is shown in blue. */
+export const TAGLINE = 'Every AI agent is a customer. FORA is your storefront.';
+
+/** "See About FORA.", linked to the page's About FORA card. */
+export const SEE_ABOUT = 'See About FORA.';
+
+/**
+ * The About FORA card (Callout kind="about"): once per page, near the end.
+ * The owner's text of the fifth round (2026-11-15), word for word.
+ */
+export const ABOUT_FORA: { title: string; text: string }[] = [
+ {
+ title: 'Federated Open Resource Access.',
+ text: 'FORA is an open protocol and free, open-source software. Any company can run its own content marketplace on it, and every FORA marketplace works with every other.',
+ },
+ {
+ title: 'Where it runs.',
+ text: 'Gutefrage runs the whole FORA stack itself on helpster.de: its own Exchange, on its own terms.',
+ },
+ {
+ title: 'Ready to try.',
+ text: 'Anyone can run FORA\'s Exchange and publisher console: Gutefrage runs its own, and the code is open to clone. Postindustria offers a ready-made deployment with a publisher console, so a publisher can try it and start selling. It charges publishers no fee beyond the payment system\'s transaction fees.',
+ },
+ {
+ title: 'Money and disputes.',
+ text: 'Money moves outside the protocol: your operator bills the buyer and pays you under its own rules. The protocol defines a dispute call, filed against a usage report, and each operator handles disputes under its own rules.',
+ },
+];
+
+/**
+ * FORA's terms, each explained once (Callout kind="define"), where the term
+ * first matters on a page. Spread one into a Callout:
+ * ``.
+ */
+export const DEFINITIONS = {
+ exchange: {
+ tag: 'What is an Exchange?',
+ text: 'An Exchange represents publishers\' catalogs. One Exchange can serve as many publishers as needed with no loss for any of them, because every URL is a separate listing with its own terms. Publishers share an Exchange when they\'re happy with how its operator settles.',
+ },
+ broker: {
+ tag: 'What is a Broker?',
+ text: 'A search service for AI agents. It asks Exchanges for their signed offers and returns them in order of relevance, optionally narrowed to the kinds of license the agent asks for. The agent decides what it buys, where and why. A Broker never buys, sells or charges, and it can\'t change an offer.',
+ },
+ operator: {
+ tag: 'What is an operator?',
+ text: 'The company that runs an Exchange, and often the publisher console that comes with it. It bills buyers and pays publishers under its own rules.',
+ },
+ edgeFunction: {
+ tag: 'What is the edge function?',
+ text: 'FORA\'s small program on your CDN, on Cloudflare, Fastly or Amazon CloudFront. On the paths you license, it refuses AI bots that come without a signed link, and serves the agents that bring one. Readers and search crawlers pass as before.',
+ },
+ signedLink: {
+ tag: 'What is a signed link?',
+ text: 'A short-lived link to one page, bound to the agent that bought it. Your CDN checks it before your site serves the page.',
+ },
+ dealTerm: {
+ tag: 'What is a deal term?',
+ text: 'For a buyer that already pays you under a contract, your Exchange can show a term priced at zero that only agents naming that deal see. The buyer still owes a usage report. The Exchange doesn\'t verify the name. A false claim is signed, so it can be traced to the key that signed it.',
+ },
+} as const;
+
+/**
+ * The search callout's FORA half. A vendor page writes one or two sentences
+ * on what the vendor does to search crawlers, then this text and these links:
+ * `…`.
+ */
+export const SEARCH_FORA = {
+ text: 'FORA\'s edge never refuses Googlebot, Bingbot or Applebot. On the paths you license it refuses AI bots that come without a signed link, AI search crawlers such as PerplexityBot included, by default. The lists are yours to change.',
+ links: [
+ { href: '/resources/ai-crawlers/', text: 'How to control which bots get in →' },
+ { href: '/components/edge-function/bot-detection/', text: 'Bot detection' },
+ ],
+};
+
+/** The comparison table's eight questions, in this order, on every vendor page. */
+export const COMPARE_QUESTIONS = [
+ 'How do you say what\'s for sale?',
+ 'How does a buyer find your content?',
+ 'Who is the buyer, provably?',
+ 'What lets a given use happen?',
+ 'Who sets the price, in what unit?',
+ 'How does money move?',
+ 'What do you see?',
+ 'What if the numbers are wrong?',
+] as const;
+
+type Eight = [T, T, T, T, T, T, T, T];
+
+/** One [vendor, FORA] pair per question, in the order of COMPARE_QUESTIONS. */
+export function compareRows(cells: Eight<[string, string]>) {
+ return COMPARE_QUESTIONS.map((q, i) => ({ q, cells: cells[i] }));
+}
+
+/** At a glance (GlanceCard): two groups, the same rows in the same order on every vendor page. */
+export const GLANCE_GROUPS = [
+ {
+ name: 'Money',
+ rows: [
+ { key: 'payer', label: 'Who pays you, and when' },
+ { key: 'fees', label: 'Fees' },
+ { key: 'reporting', label: 'What you see' },
+ { key: 'contract', label: 'Contract' },
+ { key: 'status', label: 'Status' },
+ ],
+ },
+ {
+ name: 'Your stack',
+ rows: [
+ { key: 'dns', label: 'DNS' },
+ { key: 'cdn', label: 'CDN' },
+ { key: 'firewall', label: 'Firewall and bot shield' },
+ { key: 'site', label: 'Your site and CMS' },
+ { key: 'search', label: 'Search traffic' },
+ { key: 'upkeep', label: 'Upkeep' },
+ { key: 'leaving', label: 'Leaving' },
+ ],
+ },
+] as const;
+export type GlanceKey = (typeof GLANCE_GROUPS)[number]['rows'][number]['key'];
+
+/** The GlanceCard's groups, from one [vendor, FORA] pair per row key. */
+export function glanceGroups(cells: Record) {
+ return GLANCE_GROUPS.map((g) => ({
+ name: g.name,
+ rows: g.rows.map((r) => ({ label: r.label, cells: cells[r.key] })),
+ }));
+}
+
+/** FORA's cells that read the same on every page (settled 2026-11-15; CDN, fifth round). */
+export const FORA_CELLS = {
+ /** GlanceCard, "Who pays you, and when". */
+ payer: `Your Exchange operator, under its own terms. ${SEE_ABOUT}`,
+ /** GlanceCard, "Fees". */
+ fees: `No Exchange fee at FORA's publisher console, beyond the payment system's. ${SEE_ABOUT}`,
+ /** GlanceCard, "CDN" (fifth round: a drop-in, no minutes). */
+ cdn: 'A drop-in edge function. The rest depends on your own setup; FORA adds to what you run rather than replacing it.',
+ /** GlanceCard, "Leaving". */
+ leaving: 'Delete the edge function from your CDN.',
+ /** Comparison table, "How does money move?". */
+ money: `Outside the protocol. ${SEE_ABOUT}`,
+ /** Comparison table, "What if the numbers are wrong?". */
+ disputes: `Defined in the protocol, handled under your operator's rules. ${SEE_ABOUT}`,
+};
+
+/** The series' section on the Resources index. */
+export const SERIES_SECTION = {
+ name: 'Content marketplaces and standards',
+ intro: 'Who pays publishers for AI use today, how each channel works on your site, and which standards declare what. Start with the map.',
+};
+
+export interface SeriesPage {
+ href: string;
+ title: string;
+ /** The Resources index card. */
+ summary: string;
+ audience: string;
+ formats: string[];
+ updated: string;
+ /**
+ * False until the page exists. The Resources index lists only built pages,
+ * so it never links a route the site doesn't serve; "More in this series"
+ * lists all seven, as the copy has it.
+ */
+ built: boolean;
+}
+
+/** The seven pages, in the order the index and "More in this series" list them. */
+export const SERIES: SeriesPage[] = [
+ {
+ href: '/resources/content-marketplaces/',
+ title: 'Content marketplaces and FORA',
+ summary: 'Who pays publishers for AI use of their content today, how each channel works, what your engineers change for it, and how FORA, the open protocol any company can run a marketplace on, works beside all of them.',
+ audience: 'Publishers',
+ formats: ['Guide', 'Map'],
+ updated: '2026-11-15',
+ built: true,
+ },
+ {
+ href: '/resources/content-marketplaces/tollbit/',
+ title: 'TollBit and FORA',
+ summary: 'How TollBit charges AI bots for your pages, what your engineers set up for it, and how to run both on one site.',
+ audience: 'Publishers and engineers',
+ formats: ['Guide'],
+ updated: '2026-11-15',
+ built: true,
+ },
+ {
+ href: '/resources/content-marketplaces/cloudflare/',
+ title: 'Cloudflare and FORA',
+ summary: 'How Cloudflare charges AI crawlers at its edge, what that takes on your site, and how a publisher on Cloudflare can run both.',
+ audience: 'Publishers and engineers',
+ formats: ['Guide'],
+ updated: '2026-11-15',
+ built: true,
+ },
+ {
+ href: '/resources/content-marketplaces/microsoft-pcm/',
+ title: 'Microsoft Publisher Content Marketplace and FORA',
+ summary: 'How Microsoft\'s pilot marketplace licenses content to Copilot, what it asks of your site, and how to use both.',
+ audience: 'Publishers',
+ formats: ['Guide'],
+ updated: '2026-11-15',
+ built: true,
+ },
+ {
+ href: '/resources/content-marketplaces/prorata-gist/',
+ title: 'ProRata.ai and Gist, and FORA',
+ summary: 'How ProRata pays a share of each Gist answer, what Gist Answers needs from your site, and how to run both.',
+ audience: 'Publishers',
+ formats: ['Guide'],
+ updated: '2026-11-15',
+ built: true,
+ },
+ {
+ href: '/resources/content-marketplaces/perplexity/',
+ title: "Perplexity's Publishers' Program and Comet Plus, and FORA",
+ summary: 'How Comet Plus is built to pay publishers from subscription revenue, what it asks of your site, and how to run both.',
+ audience: 'Publishers',
+ formats: ['Guide'],
+ updated: '2026-11-15',
+ built: true,
+ },
+ {
+ href: '/resources/standards-map/',
+ title: 'The standards map',
+ summary: 'Thirteen open standards tell a machine something about your content: what each answers, who uses it, how they stack, and where FORA sits.',
+ audience: 'Publishers and engineers',
+ formats: ['Guide', 'Table'],
+ updated: '2026-11-15',
+ built: true,
+ },
+];
diff --git a/website/src/data/resources.ts b/website/src/data/resources.ts
index bc679fc7..30bdfeac 100644
--- a/website/src/data/resources.ts
+++ b/website/src/data/resources.ts
@@ -1,7 +1,11 @@
// The Resources section (/resources/): one entry per topic, in the order the
// index lists them. A topic has one page that holds its article and its
// formats (slides, a generator, a video); `formats` names what that page has.
-// Topics that live elsewhere on the site link there.
+// Topics that live elsewhere on the site link there. A topic with a `section`
+// is listed under that section's heading (SECTIONS), after the topics
+// without one; a series of pages, such as the content marketplaces, is one
+// section.
+import { SERIES, SERIES_SECTION } from './content-marketplaces';
export interface Resource {
href: string;
@@ -11,8 +15,13 @@ export interface Resource {
formats: string[];
/** ISO date of the last substantive update. */
updated: string;
+ /** The name of the section the index lists it under; one of SECTIONS. */
+ section?: string;
}
+/** The index's sections, in order, each with a heading and an intro. */
+export const SECTIONS: { name: string; intro: string }[] = [SERIES_SECTION];
+
export const RESOURCES: Resource[] = [
{
href: '/resources/ai-crawlers/',
@@ -22,6 +31,22 @@ export const RESOURCES: Resource[] = [
formats: ['Guide', 'Generator'],
updated: '2026-11-12',
},
+ {
+ href: '/resources/web-bot-auth/',
+ title: 'Web Bot Auth: how an AI agent proves who it is',
+ summary: 'How an AI agent signs its requests with a key on its own website, and how your site checks it. Who signs, how keys rotate and are revoked, what to ask your CDN, and how FORA uses it.',
+ audience: 'Publishers and engineers',
+ formats: ['Guide'],
+ updated: '2026-11-16',
+ },
+ {
+ href: '/resources/agent-identity/',
+ title: 'How AI agents prove who they are, and what they may do',
+ summary: 'Four questions about any AI agent: who it is, whom it acts for, what it may do, and who pays. Which standard answers each, from Web Bot Auth and OAuth to card networks, and what you need for what you sell.',
+ audience: 'Publishers and engineers',
+ formats: ['Guide'],
+ updated: '2026-11-16',
+ },
{
href: '/resources/transparent-collective/',
title: 'A fair licensing collective',
@@ -70,6 +95,10 @@ export const RESOURCES: Resource[] = [
formats: ['Walkthrough'],
updated: '2026-11-10',
},
+ // The content-marketplaces series, in its own order; a page is listed once it is built.
+ ...SERIES.filter((p) => p.built).map(({ href, title, summary, audience, formats, updated }) => ({
+ href, title, summary, audience, formats, updated, section: SERIES_SECTION.name,
+ })),
];
/** The date a resource page shows in its hero: its entry's `updated`. */
diff --git a/website/src/lib/ai-rules.ts b/website/src/lib/ai-rules.ts
index 1eebe19a..74c60493 100644
--- a/website/src/lib/ai-rules.ts
+++ b/website/src/lib/ai-rules.ts
@@ -64,7 +64,7 @@ export function buildRules({ visible, contract, eu, termsUrl }: RuleChoices): Ru
if (contract) steps.push('Set your access fee in your terms page: paste the guide\'s wording (A), which incorporates the contract, and (B), which states your fee, and fill in the brackets. The contract itself stays unchanged: the wording to paste.');
if (!visible) steps.push('In Google Search Console, use the "Search generative AI" setting to keep your site out of AI Overviews and AI Mode. No file can set it.');
steps.push('Archive your robots.txt with the Internet Archive after each change, as dated evidence.');
- steps.push('Put FORA\'s edge function on the paths you sell. Live fetchers that ignore robots.txt get a 403 that tells them where to buy.');
+ steps.push('Put FORA\'s edge function on the paths you sell. Live fetchers that ignore robots.txt get a 403 that tells them where to buy.' + (visible ? ' By default it refuses the AI search crawlers there too. To keep those paths in AI search, let the crawlers through.' : ''));
return { robots: robots.join('\n'), headers: headers.join('\n'), html: html.join('\n'), steps };
}
diff --git a/website/src/lib/resource-markup.ts b/website/src/lib/resource-markup.ts
new file mode 100644
index 00000000..c503e664
--- /dev/null
+++ b/website/src/lib/resource-markup.ts
@@ -0,0 +1,19 @@
+// Inline markup for resource pages that also appears inside HTML strings (a
+// GlanceCard cell, a table cell, a step): a status tag and a citation marker.
+// Their styles are global, in styles/landing.css ("Resource-page atoms"), so
+// they look the same wherever a string lands. StatusTag.astro renders the same
+// markup as statusTag(), so a tag in a component and a tag in a string match.
+import { STATUS, type StatusKey } from '../data/content-marketplaces';
+
+/** A status pill: Live, Beta, Closed beta, Pilot, Preview, Experimental, Design, Roadmap, Not public. */
+export function statusTag(status: StatusKey): string {
+ return `${STATUS[status].label}`;
+}
+
+/**
+ * Citation markers: cite(1, 2) gives "[1, 2]", each number linked to its
+ * entry in the page's numbered SourceList (id="src-N").
+ */
+export function cite(...n: number[]): string {
+ return `[${n.map((i) => `${i}`).join(', ')}]`;
+}
diff --git a/website/src/pages/about.astro b/website/src/pages/about.astro
new file mode 100644
index 00000000..53f5a525
--- /dev/null
+++ b/website/src/pages/about.astro
@@ -0,0 +1,250 @@
+---
+// About FORA: why FORA exists, what it is, the company and the people who
+// build it, and how the protocol is developed. Built from the landing
+// components (src/components/landing/), like /publishers/. The positioning
+// statement and the tagline are shared with the content-marketplaces page
+// through src/data/content-marketplaces.ts. remark-standards runs on .mdx
+// only, so the standards in "How FORA is developed" are linked here by hand:
+// from src/data/standards.mjs where it lists them, and otherwise to the pages
+// the standards map links (/resources/standards-map/).
+import LandingLayout from '../layouts/LandingLayout.astro';
+import AudienceHero from '../components/landing/AudienceHero.astro';
+import Band from '../components/landing/Band.astro';
+import CardGrid from '../components/landing/CardGrid.astro';
+import PointList from '../components/landing/PointList.astro';
+import PullQuote from '../components/landing/PullQuote.astro';
+import LinkRow from '../components/landing/LinkRow.astro';
+import ClosingCta from '../components/landing/ClosingCta.astro';
+import { FORA_PUBLISHER_CONSOLE_URL } from 'astro:env/client';
+import { POSITIONING, TAGLINE } from '../data/content-marketplaces';
+import { STANDARDS } from '../data/standards.mjs';
+import { REPO_URL, BUILD_LINKS } from '../consts';
+
+const SERVER_URL = BUILD_LINKS.find((l) => l.label === 'FORA Server')!.url;
+const ISSUES_URL = `${REPO_URL}/issues`;
+const a = (href: string, text: string) => `${text}`;
+const std = (id: string, text: string) => a(STANDARDS.find((s) => s.id === id)!.url, text);
+
+// The v1 announcement on LinkedIn, 10 September 2026.
+const V1_POST = 'https://lnkd.in/p/dxJURtKj';
+
+const WHAT = [
+ {
+ title: 'An open protocol',
+ text: 'A public specification anyone can implement. Six steps take an agent from finding a resource to proving what it got: Discover, Query, Transact, Deliver, Report and Dispute.',
+ },
+ {
+ title: 'Free, open-source software',
+ text: 'FORA Server runs the whole stack: Exchange, Broker and search, identity and registration, content ingestion, an MCP endpoint and edge protection. Run all of it or one piece. SDKs for Go, TypeScript and Python.',
+ },
+ {
+ title: 'Federated, with no central registry',
+ text: 'Each marketplace, called an Exchange, runs on its own and sets its own terms. An agent registers at each Exchange it buys from, and uses the same SDK for all of them.',
+ },
+];
+
+const PEOPLE = [
+ { title: 'Constantine Mirin', text: 'CEO of Postindustria and co-author of the FORA protocol.' },
+ { title: 'Eugene Dymo', text: 'Technical Partner at Postindustria. He works on the FORA protocol and its documentation.' },
+ { title: 'Yaroslav Krasilchuk', text: 'Senior Agentic Engineer at Postindustria. He works on the FORA protocol and FORA Server.' },
+ { title: 'Serge Shtyk', text: 'Engineer. He works on FORA Server.' },
+];
+
+const OPEN = [
+ {
+ title: 'A public specification.',
+ text: `The protocol, its schemas and these docs live in one public repository, ${a(REPO_URL, 'FORA-Protocol/protocol')}. Changes land as pull requests anyone can read, and the ${a('/reference/changelog/', 'changelog')} lists every release.`,
+ },
+ {
+ title: 'A stable v1.',
+ text: 'FORA v1 was released on 10 September 2026. Within v1, fields and values can be added, and nothing already shipped is renamed, renumbered, retyped or removed. Every build checks this.',
+ },
+ {
+ title: 'Open licenses.',
+ text: `The protocol specification and FORA Server are open source under Apache 2.0, and the documentation is CC BY-ND 4.0. Deploy it without asking anyone. The ${a('/brand/', 'brand guidelines')} cover the FORA name and logo.`,
+ },
+ {
+ title: 'Built on open work.',
+ text: `FORA reads ${std('rsl', 'RSL')}, IAB Tech Lab's ${a('https://github.com/IABTechLab/CoMP', 'CoMP')}, ${a('https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/', 'TDMRep')} and ${std('c2pa', 'C2PA')}, and identifies agents with ${std('webbotauth-httpsig-protocol-00', 'Web Bot Auth')}. It is the transaction layer between them.`,
+ },
+];
+---
+
+
+ Read the protocol
+ Code on GitHub
+
+
+
+
Gutefrage runs the whole FORA stack itself, on its own servers, and has launched it on helpster.de. Its Broker searches the index Gutefrage already runs, and Gutefrage decides which fields are searched and how results rank.
+
+
+
+
+
+
The goal is for FORA to be used as widely as possible: by publishers, AI companies, marketplaces and anyone else who works with content. We help everyone who tries it, and we want everyone who builds on it to win.
+
Postindustria develops FORA, the protocol and FORA Server, and hosts a ready-made deployment with a publisher console, so a publisher can try it and start selling. It charges publishers no fee beyond the payment system's transaction fees. Anyone else can run the same: Gutefrage runs its own, and the code is open to clone.
FORA needs more than code. We want people who will use it, test it, argue with it and run it: publishers and the platforms that monetize them, AI companies building agents, content libraries and music distributors, legal and product teams, and researchers.
+
If you see a gap, a better way, or a use we haven't thought of, tell us.
Nobody browses a cable shop's website. People go where one search covers every store. That is how marketplaces won.
Now agents do the searching, and the buying, for articles, music, data and files. Selling to them will soon be as obvious as selling online.
-
Marketplaces put a middleman between you and the buyer. FORA doesn't. It is an open protocol: no one owns the customer, and your terms travel with your content.
+
No one owns your customer. FORA is an open protocol: any agent can buy from any FORA marketplace, including your own, and your terms travel with your content.
diff --git a/website/src/pages/publishers/index.astro b/website/src/pages/publishers/index.astro
index 771fbb22..32606d1c 100644
--- a/website/src/pages/publishers/index.astro
+++ b/website/src/pages/publishers/index.astro
@@ -98,7 +98,7 @@ const FAQ = [
},
{
q: 'Will I drop out of ChatGPT answers or AI Overviews?',
- a: 'No. AI Overviews, Copilot and Siri draw on Googlebot, Bingbot and Applebot, the same crawlers as ordinary search, and FORA lets them through. ChatGPT, Claude, Perplexity and Meta AI each run a search crawler of their own, and FORA lets those through too unless you choose otherwise. On your protected paths, the 403 goes to training crawlers and to bots that load a page live to answer a question: that is where an agent buys. Which bot is which.',
+ a: 'Not out of AI Overviews, Copilot or Siri. They draw on Googlebot, Bingbot and Applebot, the same crawlers as ordinary search, and FORA lets them through. ChatGPT, Claude, Perplexity and Meta AI each run a search crawler of their own. On your protected paths, FORA refuses those by default, as it refuses training crawlers and bots that load a page live to answer a question, and you can let any of them through. The rest of your site stays open to them. Which bot is which.',
},
{
q: 'What about AI training?',
diff --git a/website/src/pages/resources/agent-identity.astro b/website/src/pages/resources/agent-identity.astro
new file mode 100644
index 00000000..0b220e1e
--- /dev/null
+++ b/website/src/pages/resources/agent-identity.astro
@@ -0,0 +1,337 @@
+---
+// Resource: how AI agents prove who they are, and what they may do. The
+// companion to /resources/web-bot-auth/. Built around four questions: who the
+// agent is, whom it acts for, what it may do, and who pays. Its order: the
+// four questions, what you need for what you sell, a signature against an
+// OAuth token, OAuth and MCP, card networks, the person behind the agent,
+// anonymous credentials, the other schemes, how FORA answers the four
+// questions, the questions, and About FORA with the sources.
+//
+// Copy is written with [N] citation markers and passed through h(), which
+// turns each into cite(N), a link to entry N of the numbered SourceList;
+// SOURCES is numbered as the copy numbers it. Text with links is HTML and goes
+// through set:html.
+import LandingLayout from '../../layouts/LandingLayout.astro';
+import AudienceHero from '../../components/landing/AudienceHero.astro';
+import Band from '../../components/landing/Band.astro';
+import Callout from '../../components/landing/Callout.astro';
+import Figure from '../../components/landing/Figure.astro';
+import PullQuote from '../../components/landing/PullQuote.astro';
+import Split from '../../components/landing/Split.astro';
+import PointList from '../../components/landing/PointList.astro';
+import StepList from '../../components/landing/StepList.astro';
+import ResourceTable from '../../components/landing/ResourceTable.astro';
+import FaqGrid from '../../components/landing/FaqGrid.astro';
+import LinkRow from '../../components/landing/LinkRow.astro';
+import SourceList from '../../components/landing/SourceList.astro';
+import ClosingCta from '../../components/landing/ClosingCta.astro';
+import { resourceUpdated } from '../../data/resources';
+import { ABOUT_FORA, TAGLINE } from '../../data/content-marketplaces';
+import { cite } from '../../lib/resource-markup';
+
+const HREF = '/resources/agent-identity/';
+const IMG = '/resources/agent-identity';
+const WBA = '/resources/web-bot-auth/';
+const a = (href: string, text: string) => `${text}`;
+/** Turns each [N] or [N, M] marker in the copy into a citation link. */
+const h = (s: string) => s.replace(/\[(\d+(?:, \d+)*)\]/g, (_, g: string) => cite(...g.split(', ').map(Number)));
+
+const MASTERCARD = 'https://developer.mastercard.com/merchant-cloud/documentation/tutorials-and-guides/agentic-commerce-guide/21/';
+
+const QUESTIONS = [
+ ['Who is it?', h(`${a(WBA, 'Web Bot Auth')}: a signature checked against a key on the agent's own website. For the big crawlers, IP lists and reverse DNS [1, 2]`), h('On each request from an agent that signs. Your CDN or bot vendor can check it [3]')],
+ ['Whom does it act for?', h('A sign-in: Sign in with ChatGPT, World ID, or your own login through OAuth [5, 6, 7]'), h('When a reader\'s assistant opens the reader\'s account with you [5]')],
+ ['What may it do?', h('An OAuth token and its scopes; on FORA, the license [7, 8, 11]'), 'When you offer an API or an MCP server, or sell licenses'],
+ ['Who pays?', h('Card networks: Visa\'s Trusted Agent Protocol and Mastercard Agent Pay [15, 16]'), h('While an agent browses and when it pays; you or your CDN can check [15, 16]')],
+];
+
+const NEED = [
+ { title: 'You sell pages or licenses to AI companies.', text: 'Web Bot Auth tells you which website the buyer signs for, and the license says what it may do.' },
+ { title: 'Your subscribers\' assistants read for them.', text: 'OAuth, through your own sign-in, so each reader approves what the assistant may do.' },
+ { title: 'Agents pay you by card.', text: h('Visa\'s and Mastercard\'s agent programs, which your CDN can check [16].') },
+];
+
+const CREDENTIALS = [
+ ['What it tells you', 'Who it is and what it may do', h('What it may do, for whom, and often who it is [8, 17]'), h('Which website it signs for [1]')],
+ ['Who issues it', 'You, at sign-up', h('A sign-in server, which can be yours [7]'), h('The agent, itself [1]')],
+ ['If it\'s stolen', h('The thief is the agent [18]'), h('The same, unless the token is bound to a key [20]'), h('The thief is the agent until the key is removed and each site\'s copy expires [1]')],
+ ['Before the first request', 'A sign-up', h('Registering the app, or an app ID that is a web address, and often a person\'s approval [7, 21]'), h('Nothing by the standard. Some verifiers, Cloudflare among them, ask an agent to register its keys first [1, 3]')],
+ ['Rotation', 'Again at every site', h('At each sign-in server, or once if the app publishes its keys [21, 22]'), h('Once, in the agent\'s own key directory [1]')],
+];
+
+const MCP_STEPS = [
+ { text: h('The assistant calls your MCP server without a token. Your server refuses and says where to get one [7].') },
+ { text: h('The assistant goes to the sign-in server, which can be the one behind your subscriber login, and asks for what it needs, such as reading articles [7].') },
+ { text: h('The reader signs in there and approves [7].') },
+ { text: h('The sign-in server gives the assistant a token that works only at your server [7].') },
+ { text: h('The assistant sends the token with every request [7].') },
+];
+
+const OTHER = [
+ ['User agents and IP lists', h('Little: anyone can type a user agent, and cloud IP addresses belong to the platform [1]'), 'On every request', h('Yes, for crawlers that don\'t sign. Google still tells sites to use them [2]')],
+ ['Okta Cross App Access, Microsoft Entra Agent ID', h('An agent acting for an employee, managed in the company\'s directory [9, 10]'), 'A corporate customer\'s agents using an API you run for it', 'If you sell data or APIs to companies'],
+ ['Client certificates (mTLS)', h('The client holds a certificate [30]'), h('Between known partners. A user without one gets "an inscrutable and unskippable error" [30]'), 'No'],
+ ['Decentralized identifiers (DIDs), did:web', h('The controller of an identifier controls it "without requiring permission from any other party" [31]. did:web keeps the keys on a website, as Web Bot Auth does [32]'), 'Identity and credential programs', 'Not for bot traffic'],
+ ['Verifiable credentials', h('Claims an issuer signs about someone, "such as a driver\'s license or an education certificate" [33]'), 'Credential programs; a registry could vouch for an agent\'s operator this way', 'Not for bot traffic'],
+ ['AAuth', h('Identity and permission together, on the same kind of signature [34]'), h('Individual IETF drafts, not adopted by a working group [34]'), 'Worth watching'],
+];
+
+const FORA_ANSWERS = [
+ { title: 'Who it is.', text: h('Every FORA request is signed with Web Bot Auth, with a key the agent\'s own domain publishes [12].') },
+ { title: 'Whom it acts for.', text: h('The buyer is the agent\'s own account. An agent can label the end customer it buys for, and FORA keeps the label for attribution only [12].') },
+ { title: 'What it may do.', text: h('The license. The Exchange signs each offer, and the agent signs the offer it accepts, with the same keys as its requests [12]. The terms are those "the parties agree to when the agent accepts the offer" [11].') },
+ { title: 'Who pays.', text: 'The agent\'s account, billed by the Exchange\'s operator, which pays you under its own rules.' },
+];
+
+const FAQ = [
+ { q: 'How should I treat ChatGPT\'s agent when it acts for my subscriber?', a: h('As two things. Its signature tells you it\'s ChatGPT\'s agent [13]. Whether it may read your subscriber\'s articles takes the subscriber\'s own sign-in, through your login or Sign in with ChatGPT [5, 7].') },
+ { q: 'Is OAuth a rival to Web Bot Auth?', a: h('No. OAuth carries permission, and Web Bot Auth carries identity. The Web Bot Auth registry draft even accepts an app\'s OAuth self-description as a valid "Signature Agent Card" [4].') },
+ { q: 'Which of these can my CDN check?', a: h('Cloudflare and AWS WAF check Web Bot Auth signatures; Akamai, HUMAN and Vercel check those of ChatGPT\'s cloud browser [3, 13, 14]. Mastercard says its registry can be checked "through your edge provider" [16].') },
+];
+
+// The SourceList, numbered as the copy numbers it. Generated from the page
+// copy's source list.
+const CHECKED = '2026-11-16';
+const src = (title: string, publisher: string, href: string) => ({ title, publisher, href, checked: CHECKED });
+const SOURCES = [
+ {
+ name: "Web Bot Auth",
+ items: [
+ src("HTTP Message Signatures for automated traffic, draft-ietf-webbotauth-httpsig-protocol-01", "IETF webbotauth working group", "https://www.ietf.org/archive/id/draft-ietf-webbotauth-httpsig-protocol-01.txt"),
+ src("Authenticate requests with Web Bot Auth (experimental)", "Google for Developers", "https://developers.google.com/crawling/docs/crawlers-fetchers/web-bot-auth"),
+ src("Web Bot Auth (bot verification)", "Cloudflare docs", "https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/"),
+ src("Signature Agent Card and registry, draft-meunier-webbotauth-registry-04", "IETF, individual Internet-Draft", "https://www.ietf.org/archive/id/draft-meunier-webbotauth-registry-04.txt"),
+ ],
+ },
+ {
+ name: "Sign-in and permission",
+ items: [
+ src("Integrate Sign in with ChatGPT", "OpenAI for developers", "https://developers.openai.com/commerce/guides/integrate-sign-in-with-chatgpt"),
+ src("Agent Pass quickstart", "World", "https://docs.world.org/agent-pass/quickstart"),
+ src("Authorization, revision 2026-07-28", "Model Context Protocol", "https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization"),
+ src("RFC 6749: The OAuth 2.0 Authorization Framework", "IETF", "https://www.rfc-editor.org/rfc/rfc6749"),
+ src("Cross App Access", "Okta developer docs", "https://developer.okta.com/docs/concepts/xaa/"),
+ src("What is Microsoft Entra Agent ID?", "Microsoft Learn", "https://learn.microsoft.com/en-us/entra/agent-id/what-is-microsoft-entra-agent-id"),
+ ],
+ },
+ {
+ name: "FORA and verifiers",
+ items: [
+ src("Licensing terms", "FORA documentation", "/protocol/licensing-terms/"),
+ src("Authentication", "FORA documentation", "/protocol/authentication/"),
+ src("Help ensure cloud browser access to your site", "OpenAI for developers", "https://developers.openai.com/commerce/agent-access"),
+ src("AWS WAF Bot Control rule group", "AWS docs", "https://docs.aws.amazon.com/waf/latest/developerguide/aws-managed-rule-groups-bot.html"),
+ ],
+ },
+ {
+ name: "Payments",
+ items: [
+ src("Trusted Agent Protocol specifications", "Visa Developer", "https://developer.visa.com/docs/trusted-agent-protocol/trusted-agent-protocol-specifications"),
+ src("Agentic commerce guide", "Mastercard Developers", "https://developer.mastercard.com/merchant-cloud/documentation/tutorials-and-guides/agentic-commerce-guide/21/"),
+ ],
+ },
+ {
+ name: "Tokens",
+ items: [
+ src("RFC 9068: JSON Web Token (JWT) Profile for OAuth 2.0 Access Tokens", "IETF", "https://www.rfc-editor.org/rfc/rfc9068"),
+ src("RFC 6750: The OAuth 2.0 Authorization Framework: Bearer Token Usage", "IETF", "https://www.rfc-editor.org/rfc/rfc6750"),
+ src("RFC 9700: Best Current Practice for OAuth 2.0 Security", "IETF", "https://www.rfc-editor.org/rfc/rfc9700"),
+ src("RFC 9449: OAuth 2.0 Demonstrating Proof of Possession (DPoP)", "IETF", "https://www.rfc-editor.org/rfc/rfc9449"),
+ src("Client registration, revision 2026-07-28", "Model Context Protocol", "https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization/client-registration"),
+ src("RFC 7523: JWT Profile for OAuth 2.0 Client Authentication and Authorization Grants", "IETF", "https://www.rfc-editor.org/rfc/rfc7523"),
+ src("Introduction", "Eclipse Biscuit documentation", "https://doc.biscuitsec.org/getting-started/introduction"),
+ ],
+ },
+ {
+ name: "People and privacy",
+ items: [
+ src("ID tokens and identity", "World", "https://docs.world.org/agent-pass/id-tokens-and-identity"),
+ src("Use World ID with your agent", "World", "https://docs.world.org/agent-pass/use-world-id-with-your-agent"),
+ src("Hosted Key Directories for Web Bot Auth, draft-singh-webbotauth-hosted-directories-00", "IETF, individual Internet-Draft", "https://datatracker.ietf.org/doc/draft-singh-webbotauth-hosted-directories/"),
+ src("Anonymous credentials: rate-limiting bots and agents without compromising privacy", "Cloudflare blog", "https://blog.cloudflare.com/private-rate-limiting/"),
+ src("Private Access Tokens challenge", "Apple Developer", "https://developer.apple.com/news/?id=huqjyh7k"),
+ src("WebBotAuth Direction", "the working group's chairs, IETF web-bot-auth mailing list", "https://mailarchive.ietf.org/arch/msg/web-bot-auth/gxkMGMC75Sm1LO9WPvZbPKYhNAc/"),
+ ],
+ },
+ {
+ name: "Other schemes",
+ items: [
+ src("Forget IPs: using cryptography to verify bot and agent traffic", "Cloudflare blog", "https://blog.cloudflare.com/web-bot-auth/"),
+ src("Decentralized Identifiers (DIDs) v1.0", "W3C Recommendation", "https://www.w3.org/TR/did-1.0/"),
+ src("did:web Method Specification", "W3C Credentials Community Group", "https://w3c-ccg.github.io/did-method-web/"),
+ src("Verifiable Credentials Data Model v2.0", "W3C Recommendation", "https://www.w3.org/TR/vc-data-model-2.0/"),
+ src("AAuth, draft-hardt-oauth-aauth-protocol", "IETF, individual Internet-Draft", "https://datatracker.ietf.org/doc/draft-hardt-oauth-aauth-protocol/"),
+ ],
+ },
+];
+
+const MORE = [
+ { href: WBA, text: 'Web Bot Auth: how an AI agent proves who it is' },
+ { href: '/resources/standards-map/', text: 'The standards map' },
+ { href: '/resources/content-marketplaces/', text: 'Content marketplaces and FORA' },
+];
+---
+
+
+ The four questions
+ What you need
+
+
+
+
+
+
+
+
+
+
+
+
The rest of this page you can leave to your engineers for now.
+
+
+
+
+
+ See your offers
+ How FORA agents sign
+
+
+
+
+
+
diff --git a/website/src/pages/resources/ai-crawlers.astro b/website/src/pages/resources/ai-crawlers.astro
index 1e7549ef..f77cf92f 100644
--- a/website/src/pages/resources/ai-crawlers.astro
+++ b/website/src/pages/resources/ai-crawlers.astro
@@ -54,7 +54,7 @@ const FAQ = [
},
{
q: 'What about Perplexity?',
- a: 'Perplexity says PerplexityBot builds its search index and is not used to train models, so refusing it takes you out of Perplexity\'s results. Perplexity-User loads pages when someone asks and generally ignores robots.txt: on the paths you sell, FORA gives it a 403.',
+ a: 'Perplexity says PerplexityBot builds its search index and is not used to train models, so refusing it takes you out of Perplexity\'s results. On the paths you sell, FORA\'s edge function refuses it by default, and you can let it through. Perplexity-User loads pages when someone asks and generally ignores robots.txt: on the paths you sell, FORA gives it a 403.',
},
{
q: 'Does robots.txt bind anyone?',
@@ -99,7 +99,7 @@ const FAQ = [
title="Each AI company reads its own refusal"
lead="The AI companies treat their uses as allowed unless a site refuses them, and there is no single signal they all honor. So a refusal has to be said in each vendor's own form, in up to four places."
>
-
+
-
+ `${text}`;
+
+const GLANCE = glanceGroups({
+ payer: [`Cloudflare, as merchant of record, monthly, to a new Stripe Connect account, after settlement periods and payout thresholds ${cite(1, 2)}.`, FORA_CELLS.payer],
+ fees: ['Not public. Cloudflare\'s docs, posts and terms state no fee.', FORA_CELLS.fees],
+ reporting: [
+ `Which crawlers visit and how often they're charged. Your earnings balance isn't in the dashboard ${cite(3)}.`,
+ `Each sale with its usage-report status, and daily income with a CSV download, in your operator's publisher console ${cite(33)}.`,
+ ],
+ contract: [
+ `Pay Per Crawl's program terms: not public. Pay Per Use: each program's terms say what the buyer may do ${cite(4)}. No dispute process documented.`,
+ `Your agreement with your Exchange operator. Your terms go into each offer the Exchange signs ${cite(34)}. Disputes: see About FORA.`,
+ ],
+ status: [
+ `Pay Per Crawl ${statusTag('closed-beta')}, charging listed under Enterprise plans with Bot Management ${cite(1, 5)}. Pay Per Use ${statusTag('beta')} ${cite(4)}. No AI company named as paying through Pay Per Crawl.`,
+ `${statusTag('live')} ${SEE_ABOUT}`,
+ ],
+ dns: [
+ `Cloudflare's nameservers, with DNSSEC off first, or your own DNS provider on Business or Enterprise ${cite(6, 7, 8)}. Every web hostname routed through Cloudflare ${cite(9)}.`,
+ `No record to add ${cite(33)}.`,
+ ],
+ cdn: [
+ `Cloudflare only ${cite(6)}.`,
+ 'A drop-in edge function. The rest depends on your own setup; FORA adds to what you run rather than replacing it.',
+ ],
+ firewall: [
+ `Your rules run first, so a crawler they block is never offered a price ${cite(10)}. Turn off Block AI Bots to join ${cite(11)}.`,
+ `Your WAF and bot product run first. Let signed requests and your two files through them; FORA's guide gives two setups ${cite(35)}.`,
+ ],
+ site: [
+ `No change for one price. A price per page needs a response header from your origin or a Worker ${cite(12)}.`,
+ `One file, fora.json, naming your Exchanges, and a sitemap the operator imports your pages from ${cite(33)}.`,
+ ],
+ search: [
+ `Only crawlers you set to Charge pay. Cloudflare warns that charging a search crawler may hurt your SEO ${cite(13)}.`,
+ `Googlebot, Bingbot and Applebot always pass ${cite(35)}. See Search traffic, below.`,
+ ],
+ upkeep: [
+ `Cloudflare keeps the list of crawlers you can charge ${cite(14)}. Logpush can't fill in the past, and logs are lost while a job fails ${cite(15)}.`,
+ `Change your Exchanges by editing your discovery file, with nothing to redeploy ${cite(33)}.`,
+ ],
+ leaving: ['The docs describe no step to turn Pay Per Crawl off, close its Stripe account or take a final payout.', FORA_CELLS.leaving],
+});
+
+const SHORT_ANSWER = [
+ { title: 'Cloudflare charges at the door.', text: 'Pay Per Crawl charges a crawler one price per page, and Cloudflare collects and pays out. In Pay Per Use, an AI company proposes a price for a use, and you accept or decline. Both are in beta. Pay Per Crawl needs your site\'s traffic to run through Cloudflare.' },
+ { title: 'FORA sells the license.', text: 'You set the uses, buyers, caps and price for each resource, and can offer one article on several terms. An AI agent (software that acts for a user, one task at a time) buys under an offer your Exchange signs. Pay Per Crawl needs no Exchange of yours, because Cloudflare is one.' },
+ { status: 'closed-beta' as const, title: 'Pay Per Crawl is by admission only.', text: 'Cloudflare\'s plan table lists charging under Enterprise plans with Bot Management. You need a new Stripe account, and your balance isn\'t in the dashboard. Cloudflare names no AI company paying through it.' },
+];
+
+const PPC_STEPS = [
+ { title: 'Your site runs through Cloudflare, and is admitted.', text: `"Your domain must be using Cloudflare's nameservers, or have DNS records managed by Cloudflare" (${a(C.enable, 'Enable in account settings')}). The program is a closed beta (${a(C.whatIs, 'What is Pay Per Crawl')}). The plan table lists charging under Enterprise plans with Bot Management (${a(C.getStarted, 'Get started')}).` },
+ { title: 'You set one price and pick who pays it.', text: `Cloudflare lists the AI crawlers it knows, and you set each one to Charge, Allow or Block. Cloudflare says "you can only set a single price" for every crawler set to Charge (${a(C.faq, 'FAQ')}), and "The minimum price is $0.001 USD per crawl" (${a(C.setPrice, 'Set a price')}).` },
+ { title: 'The crawler registers.', text: `It signs its requests with Web Bot Auth, an IETF draft for signing bot requests. It passes Cloudflare's review and connects Stripe (${a(C.crawlPages, 'Crawl pages')}).` },
+ { title: 'Cloudflare quotes.', text: 'An unpaid request gets HTTP 402, Payment Required, with the price. The crawler retries with a signed agreement to it, or states a maximum up front.' },
+ { title: 'Cloudflare charges each successful fetch.', text: `"Error responses are not billed", and a crawler "will incur the cost set by the site owner" on every visit (${a(C.faq, 'FAQ')}).` },
+ { title: 'Cloudflare pays you.', text: `"Cloudflare acts as the Merchant of Record for pay per crawl" (${a(C.whatIs, 'What is Pay Per Crawl')}): it is the party that takes the crawler's payment. It pays monthly to a dedicated Stripe Connect account, "subject to settlement periods and minimum payout thresholds" (${a(C.payouts, 'Manage payouts')}).` },
+];
+
+const KEY_FACTS = [
+ { value: '$0.001', text: `The lowest price you can set per crawl ${cite(17)}` },
+ { value: '1', text: `Price for every crawler you set to Charge, unless your origin or a Worker prices each page ${cite(12, 16)}` },
+ { value: '20%+', text: `Of the web behind Cloudflare's network, the vendor's own figure ${cite(19)}` },
+];
+
+const SETUP = [
+ { title: 'Your DNS and proxy.', text: `Your domain moves to Cloudflare's nameservers, and "If your domain uses DNSSEC, you must turn it off at your registrar before changing nameservers" (${a(C.onboard, 'Onboard a domain')}). Keeping your DNS provider is "only available to customers on a Business or Enterprise plan" (${a(C.cname, 'CNAME setup')}). A hostname not routed through Cloudflare (left unproxied) never meets Pay Per Crawl (${a(C.proxy, 'Proxy status')}).` },
+ { title: 'Your content stays on your servers.', text: `Cloudflare keeps no copy or catalog of your pages. A crawler gets the list of domains that charge it and finds the pages itself (${a(C.discover, 'Discover payable content')}). "The Cloudflare CDN does not cache HTML or JSON by default", so each crawl normally reaches your origin (${a(C.cache, 'Default cache behavior')}).` },
+ { title: 'Cloudflare decides which bots you can charge.', text: `You choose Charge, Allow or Block only for crawlers on Cloudflare's list. Bot operators apply to join it, and "Cloudflare reserves the right to re-assign Verified bot categories" (${a(C.verifiedBots, 'Verified bots')}). On the free plan, AI Crawl Control "identifies AI crawlers based on their user agent strings" (${a(C.manage, 'Manage AI crawlers')}).` },
+ { title: 'Your other rules run first.', text: `Firewall rules and bot tools act before Pay Per Crawl, so a crawler they block is never offered a price (${a(C.waf, 'AI Crawl Control with WAF')}).` },
+ { title: 'Per-page prices need a header.', text: `Since June 2026, your origin or a Worker can price each page in a response header (${a(C.advanced, 'Advanced configuration')}).` },
+ { title: 'Refusals and redirects.', text: `A blocked crawler gets a 403 (refused), or on paid plans a 402 (pay first) with a message of your own (${a(C.manage, 'Manage AI crawlers')}). The one redirect shipped sends a verified AI crawler to the page's canonical URL (${a(C.redirects, 'Redirects for AI Training')}), not to a license or a marketplace.` },
+ { title: 'Logs.', text: `Cloudflare's log export, Logpush, streams each request with a Pay Per Crawl outcome, whose values aren't listed (${a(C.logpush, 'Logpush')}).` },
+ { title: 'FORA agents and Cloudflare\'s signature check.', text: `Cloudflare's verifier reads an older form of Web Bot Auth than the one FORA's software sends, and "will fail to verify a message if" it uses the later form (${a(C.webBotAuth, 'Web Bot Auth')}). So Cloudflare can't verify a FORA agent by its signature, and the agent can't pay Pay Per Crawl with it. FORA sales don't depend on it: the agent brings a signed link, which FORA's Worker checks. Let signed requests through your rules, as under Running both. The working group's -01, of 2026-10-29, lets a verifier accept Cloudflare's older form and tells signers to send the form FORA's software sends (${a(C.draft01, 'draft-ietf-webbotauth-httpsig-protocol-01')}).` },
+];
+
+const CAUTIONS = [
+ { title: 'Your balance comes on request.', text: `"You can request balance updates from your Cloudflare team" (${a(C.monitor, 'Monitor activity')}).` },
+ { title: 'You need a new Stripe account.', text: `"Pre-existing Stripe accounts are not compatible with this feature" (${a(C.payouts, 'Manage payouts')}).` },
+ { title: 'Logs can be lost.', text: `"If your job is disabled or fails, logs generated during that period are permanently lost" (${a(C.logpush, 'Logpush')}). Logpush is billed per gigabyte above 25 GB a month on the Free to Business plans (${a(C.logpushPricing, 'Logpush pricing')}).` },
+ { title: 'Country blocks win.', text: `Crawlers from countries you block "will continue to be blocked, even if they provide the required headers" (${a(C.waf, 'AI Crawl Control with WAF')}).` },
+ { title: 'A bad price is your error.', text: `An invalid per-page price reaches the crawler as an error: "This is a site owner configuration issue" (${a(C.errors, 'Error codes')}).` },
+];
+
+const AHEAD_FIRST = [
+ { title: 'It enforces where FORA can\'t.', text: `"More than 20% of the web sits behind Cloudflare's network" (${a(C.smarter, 'Making AI search smarter')}, published 2026-07-01; vendor's own figure). A crawler you set to Charge with no billing relationship gets "the functional equivalent of a network level block" (${a(C.launch, 'launch post')}, published 2025-07-01). In July 2026, Cloudflare announced that on 2026-09-15 it would change the settings of every free customer that hadn't changed them (${a(C.press, 'press release')}). FORA's edge function refuses unlicensed AI bots only on the sites that install it.` },
+];
+const AHEAD_REST = [
+ { title: 'It detects bots, and FORA uses that.', text: `Cloudflare verifies a bot by its signature, a published IP list or reverse DNS (${a(C.verifiedBots, 'Verified bots')}).` },
+ { title: 'It moves the money.', text: `Cloudflare is the merchant of record ${cite(1)}. In 2025 it said beta crawlers are "charged a single fee via a connected credit card or bank account at the end of each day" (${a(C.x402, 'x402 post')}, published 2025-09-23). FORA never moves money, by design. ${SEE_ABOUT}` },
+ { title: 'The price travels in HTTP.', text: `A registered crawler reads the price from a 402 response and agrees to it with one signed header (${a(C.crawlPages, 'Crawl pages')}). A FORA purchase takes protocol calls, in practice through an SDK or FORA Server's Identity Service.` },
+ { title: 'No code, on a site already on Cloudflare.', text: `There, one price for your site takes dashboard settings, admission to the beta, a review of your security rules and a new Stripe account. Pay Per Use needs "no origin change" ${cite(4)}. FORA needs a fora.json file, the edge function and an Exchange.` },
+];
+
+const COMPARE = compareRows([
+ [
+ `Dashboard settings: allow, charge or block per listed crawler, one price, free paths ${cite(12, 13, 16)}. Preferences go in robots.txt as Content Signals, a line that states a preference with no price ${cite(30)}. Cloudflare's docs describe no license in the 402.`,
+ `A file on your site names your Exchanges ${cite(34)}. Each resource can carry several terms, up to 32: uses, buyers, places, caps, duties and a price ${cite(36)}.`,
+ ],
+ [
+ `A registered crawler asks Cloudflare for the domains that charge it, finds pages itself, and gets each page's price from a 402 ${cite(18, 20)}. Pay Per Use: the domains that accepted its offer ${cite(4)}.`,
+ `An agent asks your Exchange, or a Broker, which returns the signed offers whole (${a('/resources/content-marketplaces/#broker-order', 'how a Broker orders offers')}) ${cite(37)}.`,
+ ],
+ [
+ `A Web Bot Auth signature in an older draft's form, which deployed crawlers still send, plus a reviewed registration ${cite(18, 24, 40)}. On the free plan, the user agent ${cite(22)}.`,
+ `A Web Bot Auth signature in the form of the working group's first draft (-00), checked against a key on the buyer's domain ${cite(38)}.`,
+ ],
+ [
+ `Cloudflare's edge, per request: a verified crawler set to Charge, with a signed price ${cite(18)}. Use limits sit in robots.txt, or in each Pay Per Use program's terms ${cite(3, 4)}.`,
+ `An offer the Exchange signs and the agent accepts and signs, then a signed link that expires, checked at your CDN ${cite(34)}.`,
+ ],
+ [
+ `You: one price per fetch for every charged crawler, or a price per page ${cite(12, 16)}. Pay Per Use: the buyer sets a price per use; you accept or decline ${cite(4)}.`,
+ `You, per term: free, per unit (a fetch, an access, a token) or flat. One article can be free for research and paid for commercial use. No revenue-share price ${cite(36)}.`,
+ ],
+ [`Cloudflare charges the buyer through Stripe and pays you monthly ${cite(2)}. Its fee isn't published.`, FORA_CELLS.money],
+ [
+ `Which crawlers visit, how often they're charged, robots.txt violations. "Your accrued earnings balance is not currently visible in the dashboard." (${a(C.monitor, 'Monitor activity')})`,
+ `A signed record of each sale, with its usage-report status, in the publisher console your Exchange operator runs ${cite(34)}. No protocol call returns your sales.`,
+ ],
+ [`No dispute process is documented. Pay Per Use reports are the buyer's own; Cloudflare checks each names an enrolled site ${cite(4)}.`, FORA_CELLS.disputes],
+]);
+
+const WORKS = [
+ { title: 'FORA\'s edge runs on Cloudflare.', text: `On Cloudflare, FORA's edge function is a Worker on a route that covers your hostname. It gets both Cloudflare's detection, through its verified-bot signal, and FORA's own lists, and the deployment is yours. Cloudflare's protections, AI Crawl Control included, run first, so let signed requests, your FORA and RSL files, and the Exchange's reads of your sitemap through them.FORA's guide gives two setups: the edge function decides for every client on the paths you license, or your bot product stays in front and signed requests skip it (${a('/components/edge-function/bot-detection/#cloudflare', 'Bot detection')}). The edge function also runs on CloudFront, so FORA doesn't need Cloudflare's DNS.` },
+ { title: 'Point every crawler at your license.', text: `Cloudflare documents a rule that adds a link to your terms or RSL license to its responses, "even when those crawlers are blocked" (${a(C.transform, 'Transform Rules')}).` },
+ { title: 'Keep your signals and terms in step.', text: `Cloudflare's Markdown for Agents adds a signal allowing training when your origin sends none (${a(C.markdown, 'Markdown for Agents')}). If a FORA term forbids training, send a signal that says so.` },
+];
+
+const TODO = [
+ { title: 'Decide per crawler first.', text: `AI Crawl Control allows or blocks each crawler on every plan, with no beta to wait for (${a(C.getStarted, 'Get started')}).` },
+ { title: 'Check your Cloudflare setup before you join.', text: `Every web hostname must run through Cloudflare's proxy. List the firewall and bot rules that would block a crawler you want to charge. Pay Per Crawl runs after Cloudflare's bot tools, so "you need to first turn off Block AI Bots" (${a(C.bots, 'AI Crawl Control with bots')}).` },
+ { title: 'Ask before you join Pay Per Crawl.', text: 'Ask for the fee, the payout threshold and which crawlers pay; none is published. FORA itself is free, open-source software. Agree how often you\'ll get balance updates, and plan for a new Stripe account. For the terms and the exit, see For legal.' },
+ { title: 'Start your logs on day one.', text: 'Logpush can\'t fill in the past, so set it up before the first charge.' },
+ { title: 'Sell to agents as well.', text: 'Name your FORA Exchanges in fora.json, set terms per resource, and run FORA\'s edge on the paths you license.' },
+];
+
+const LEGAL = [
+ { title: 'Pay Per Crawl\'s terms aren\'t published.', text: 'Ask for the program terms before you join. Cloudflare is the merchant of record.' },
+ { title: 'Pay Per Use sets use limits per program.', text: `Each program's terms define "what the AI company may do with the content, including any restrictions on training" (${a(C.payPerUse, 'Pay Per Use')}). The buyer reports its own uses.` },
+ { title: 'Leaving.', text: 'The docs describe no step to turn Pay Per Crawl off, close its Stripe account or take a final payout.' },
+ { title: 'Disputes.', text: 'Neither program documents a dispute process.' },
+];
+
+// The SourceList, numbered as the copy numbers it: an entry's position,
+// counted across the groups, is the N in cite(N).
+const cf = (title: string, href: string, checked = '2026-11-15') => ({ title, publisher: 'Cloudflare', href, checked });
+const fora = (title: string, href: string) => ({ title, publisher: 'FORA documentation', href, checked: '2026-11-15' });
+const SOURCES = [
+ {
+ name: 'Cloudflare',
+ items: [
+ cf('What is Pay Per Crawl?', C.whatIs),
+ cf('Manage payouts', C.payouts),
+ cf('Monitor activity', C.monitor),
+ cf('Pay Per Use', C.payPerUse, '2026-11-14'),
+ cf('Get started with AI Crawl Control', C.getStarted),
+ cf('Enable in account settings', C.enable),
+ cf('Onboard a domain', C.onboard),
+ cf('CNAME setup', C.cname),
+ cf('Proxy status', C.proxy),
+ cf('AI Crawl Control with WAF', C.waf),
+ cf('AI Crawl Control with Bots', C.bots),
+ cf('Advanced configuration', C.advanced),
+ cf('Select crawlers to charge', C.selectCrawlers),
+ cf('Verified bots', C.verifiedBots),
+ cf('Logpush', C.logpush),
+ cf('Pay Per Crawl FAQ', C.faq),
+ cf('Set a pay per crawl price', C.setPrice),
+ cf('Crawl pages', C.crawlPages),
+ cf('Making AI search smarter', C.smarter, '2026-11-14'),
+ cf('Discover payable content', C.discover),
+ cf('Default cache behavior', C.cache),
+ cf('Manage AI crawlers', C.manage),
+ cf('Redirects for AI Training', C.redirects),
+ cf('Web Bot Auth', C.webBotAuth),
+ cf('Logpush pricing', C.logpushPricing),
+ cf('Error codes', C.errors),
+ cf('Introducing pay per crawl', C.launch, '2026-11-14'),
+ cf('Cloudflare allows the agentic internet to flourish with a simple philosophy: your content, your rules', C.press, '2026-11-14'),
+ cf('x402', C.x402, '2026-11-14'),
+ cf('Managed robots.txt', C.robots),
+ cf('AI Crawl Control with Transform Rules', C.transform),
+ cf('Markdown for Agents', C.markdown),
+ ],
+ },
+ {
+ name: 'FORA',
+ items: [
+ fora('Publisher onboarding', '/getting-started/publisher-onboarding/'),
+ fora('What is FORA?', '/getting-started/what-is-fora/'),
+ fora('Bot detection', '/components/edge-function/bot-detection/'),
+ fora('Licensing terms', '/protocol/licensing-terms/'),
+ fora('The Broker', '/components/broker/overview/'),
+ fora('Authentication', '/protocol/authentication/'),
+ ],
+ },
+ {
+ name: 'Press, filings and others',
+ items: [
+ { title: 'Cloudflare says bot blocking is fueling publisher AI deals', publisher: 'Press Gazette', href: C.pressGazette, checked: '2026-11-14' },
+ { title: 'HTTP Message Signatures for automated traffic (draft-ietf-webbotauth-httpsig-protocol-01)', publisher: 'IETF', href: C.draft01, checked: '2026-11-14' },
+ ],
+ },
+];
+---
+
+
+ Where Cloudflare is ahead
+ Run both
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Pay Per Crawl, in six steps:
+
+
Pay Per Use reverses who sets the price (${a(C.payPerUse, 'Pay Per Use')}, published 2026-09-30). An AI company "defines the use it will pay for and sets a price", such as returning an excerpt of your page to a customer. You accept offers in the dashboard, with "no origin change". "Usage is self-reported": the buyer reports its own uses, and Cloudflare charges the buyer and pays you monthly. Its reason for the program: "AI products fetch far more than they use".`} />
+
Settling a FORA sale through Pay Per Crawl. Pay Per Crawl charges when a crawler fetches a page. Its docs describe no way to bill an amount for a sale made elsewhere.
+
A Cloudflare price a FORA agent can verify. Cloudflare's docs describe no signed price or receipt. The crawler gets a header that states the amount charged.
+
+
+
+
+
+
+
+
+
+
+
Sources
+
Sources are checked on the date of publication.
+
+
+
+
+
+
+ What is FORA?
+ Run your own marketplace
+ Licensing terms
+ All content marketplaces
+
+
+
+
diff --git a/website/src/pages/resources/content-marketplaces/index.astro b/website/src/pages/resources/content-marketplaces/index.astro
new file mode 100644
index 00000000..1c8f707a
--- /dev/null
+++ b/website/src/pages/resources/content-marketplaces/index.astro
@@ -0,0 +1,988 @@
+---
+// Resource: Content marketplaces and FORA, the hub of the content-marketplaces
+// series. The five deep pages sit under it at
+// /resources/content-marketplaces//. In the hub's own order: hero, the
+// short answer, what FORA takes (FORA's half of the GlanceCard, as a pair),
+// who has shown money moving (KeyFacts), the map (MarketMap), how each kind
+// works, where they are ahead, the eight questions, what changes on your CDN
+// (id="together"), when an AI company buys (with id="broker-order", which the
+// deep pages link), what to do, the deep pages, and About FORA with the
+// sources and the series navigation.
+//
+// Shared texts (About FORA, the definitions, the search callout, the glance
+// row labels and FORA's fixed cells) come from
+// src/data/content-marketplaces.ts. A claim cites its source inline, as a
+// link, or with cite(N), which links to entry N of the numbered SourceList at
+// the foot. SOURCES below is numbered in order of first citation, as the copy
+// numbers it, so an entry's position is its number; src(N) gives entry N's
+// URL for an inline link. Text with links is HTML and goes through set:html.
+import LandingLayout from '../../../layouts/LandingLayout.astro';
+import AudienceHero from '../../../components/landing/AudienceHero.astro';
+import Band from '../../../components/landing/Band.astro';
+import CardGrid from '../../../components/landing/CardGrid.astro';
+import Callout from '../../../components/landing/Callout.astro';
+import ForaGlance from '../../../components/landing/ForaGlance.astro';
+import KeyFacts from '../../../components/landing/KeyFacts.astro';
+import MarketMap from '../../../components/landing/MarketMap.astro';
+import MoneyTiles from '../../../components/landing/MoneyTiles.astro';
+import PullQuote from '../../../components/landing/PullQuote.astro';
+import PointList from '../../../components/landing/PointList.astro';
+import StepList from '../../../components/landing/StepList.astro';
+import Figure from '../../../components/landing/Figure.astro';
+import Split from '../../../components/landing/Split.astro';
+import StatusTag from '../../../components/landing/StatusTag.astro';
+import SourceList from '../../../components/landing/SourceList.astro';
+import SeriesNav from '../../../components/landing/SeriesNav.astro';
+import ClosingCta from '../../../components/landing/ClosingCta.astro';
+import { RESOURCES, resourceUpdated } from '../../../data/resources';
+import {
+ ABOUT_FORA, DEFINITIONS, SEARCH_FORA, FORA_CELLS, SEE_ABOUT, GLANCE_GROUPS, SERIES, POSITIONING, TAGLINE, type GlanceKey,
+} from '../../../data/content-marketplaces';
+import { cite, statusTag } from '../../../lib/resource-markup';
+
+const HREF = '/resources/content-marketplaces/';
+const IMG = '/resources/content-marketplaces/hub';
+const a = (href: string, text: string) => `${text}`;
+const ABOUT = 'About FORA';
+
+// The SourceList, numbered in order of first citation: an entry's position,
+// counted across the groups, is the N in cite(N).
+const SOURCES = [
+ {
+ name: 'FORA',
+ items: [
+ { title: 'Publisher onboarding', publisher: 'FORA documentation', href: '/getting-started/publisher-onboarding/', checked: '2026-11-15' },
+ { title: 'Edge function deployment', publisher: 'FORA documentation', href: '/components/edge-function/deployment/', checked: '2026-11-15' },
+ { title: 'Bot detection', publisher: 'FORA documentation', href: '/components/edge-function/bot-detection/', checked: '2026-11-15' },
+ { title: 'What is FORA?', publisher: 'FORA documentation', href: '/getting-started/what-is-fora/', checked: '2026-11-15' },
+ { title: 'Licensing terms', publisher: 'FORA documentation', href: '/protocol/licensing-terms/', checked: '2026-11-15' },
+ { title: 'Identity Service', publisher: 'FORA documentation', href: '/components/identity/overview/', checked: '2026-11-15' },
+ { title: 'Broker', publisher: 'FORA documentation', href: '/components/broker/overview/', checked: '2026-11-15' },
+ { title: 'For content marketplaces', publisher: 'FORA documentation', href: '/getting-started/existing-marketplace/', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'AI companies buying directly',
+ items: [
+ { title: '400 publications are now partnering with us on our News AI pilot program', publisher: 'Google', href: 'https://blog.google/products-and-platforms/products/news/news-ai-pilot-update-publishers/', checked: '2026-11-15' },
+ { title: 'Reddit, Inc. Form S-1', publisher: 'U.S. SEC, EDGAR', href: 'https://www.sec.gov/Archives/edgar/data/1713445/000162828024006294/reddits-1q423.htm', checked: '2026-11-14' },
+ { title: 'Google Search Console testing paying publishers for use of content in AI: AI Contribution Pilot', publisher: 'Search Engine Roundtable', href: 'https://www.seroundtable.com/google-al-contribution-pilot-42076.html', checked: '2026-11-15' },
+ { title: 'The Extended News Previews Program', publisher: 'Google Search Console Help', href: 'https://support.google.com/webmasters/answer/11317369?hl=en', checked: '2026-11-15' },
+ { title: 'Google is paying 100 publishers for AI answers, The Information reports', publisher: 'TNW', href: 'https://thenextweb.com/news/google-ai-contribution-pilot-publishers-payments', checked: '2026-11-15' },
+ { title: 'Building toward a sustainable content economy for the agentic web', publisher: 'Microsoft Advertising', href: 'https://about.ads.microsoft.com/en/blog/post/february-2026/building-toward-a-sustainable-content-economy-for-the-agentic-web', checked: '2026-11-15' },
+ { title: 'Who\'s suing AI and who\'s signing', publisher: 'Press Gazette', href: 'https://pressgazette.co.uk/platforms/news-publisher-ai-deals-lawsuits-openai-google/', checked: '2026-11-14' },
+ { title: 'Anthropic strikes its first licensing deal, with the AP, for election coverage', publisher: 'Adweek', href: 'https://www.adweek.com/media/anthropic-licensing-deal-ap-election/', checked: '2026-11-14' },
+ { title: 'OpenAI, Grupo Folha and Grupo UOL announce strategic content partnership', publisher: 'OpenAI', href: 'https://openai.com/index/grupo-folha-grupo-uol-partnership/', checked: '2026-11-14' },
+ { title: 'OpenAI signs first Brazilian content licensing deals with Folha and UOL', publisher: 'LatAm Journalism Review', href: 'https://latamjournalismreview.org/news/openai-signs-first-brazilian-content-licensing-deals-with-folha-and-uol/', checked: '2026-11-14' },
+ { title: 'How HTTP status codes affect Google\'s crawlers', publisher: 'Google Search Central', href: 'https://developers.google.com/crawling/docs/troubleshooting/http-status-codes', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Collectives and rights agencies',
+ items: [
+ { title: 'Agreement on the use of copyright-protected material in the training of Norwegian language models', publisher: 'Kopinor', href: 'https://kopinor.no/wp-content/uploads/2026/02/Agreement-on-the-use-of-copyright-protected-material-in-the-training-of-Norwegian-language-models.pdf', checked: '2026-11-14' },
+ { title: 'Collective licensing for the AI-first internet', publisher: 'Internet Licensing Collective', href: 'https://internetcollective.org/', checked: '2026-11-14' },
+ { title: 'AI Systems Training License: frequently asked questions (FY26)', publisher: 'CCC', href: 'https://www.copyright.com/crc/wp-content/uploads/sites/2/FY26-FAQ-ASTL-A4.pdf', checked: '2026-11-15' },
+ { title: 'CCC launching new AI content re-use rights for U.S. academic customers and transactional licensing capabilities for AI', publisher: 'CCC', href: 'https://www.copyright.com/media-press-releases/ccc-launching-new-ai-content-re-use-rights-for-u-s-academic-customers-and-transactional-licensing-capabilities-for-ai/', checked: '2026-11-15' },
+ { title: 'CCC Marketplace: title search and permission request', publisher: 'CCC', href: 'https://marketplace.copyright.com/rs-ui-web/mp/search', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Bot paywalls at your CDN',
+ items: [
+ { title: 'The Washington Post\'s Arc XP adds TollBit to help publishers make money from AI bot traffic', publisher: 'Digiday', href: 'https://digiday.com/media/the-washington-posts-arc-xp-adds-tollbit-to-help-publishers-make-money-from-ai-bot-traffic/', checked: '2026-11-14' },
+ { title: 'Content License Agreement (standard license)', publisher: 'TollBit', href: 'https://gateway.tollbit.com/license/j4ikxb8qqsu2qu91q78a9e70/ON_DEMAND_LICENSE_ka02vqvjacoy2i5vs8z90hxm', checked: '2026-11-15' },
+ { title: 'Setting rates', publisher: 'TollBit docs', href: 'https://docs.tollbit.com/docs/setting-rates', checked: '2026-11-14' },
+ { title: 'Platform overview', publisher: 'TollBit', href: 'https://tollbit.com/docs/', checked: '2026-11-15' },
+ { title: 'Announcing our Series A', publisher: 'TollBit blog', href: 'https://tollbit.com/blog/series-a/', checked: '2026-11-14' },
+ { title: 'TNL Mediagene announces early success in AI content licensing revenue model via TollBit marketplace integration and strategic partnership', publisher: 'PR Newswire', href: 'https://www.prnewswire.com/news-releases/tnl-mediagene-announces-early-success-in-ai-content-licensing-revenue-model-via-tollbit-marketplace-integration-and-strategic-partnership-302647966.html', checked: '2026-11-14' },
+ { title: 'Enable analytics', publisher: 'TollBit docs', href: 'https://docs.tollbit.com/docs/step-2-enable-analytics', checked: '2026-11-15' },
+ { title: 'Platform architecture', publisher: 'TollBit docs', href: 'https://docs.tollbit.com/docs/platform-architecture', checked: '2026-11-14' },
+ { title: 'Verifying a property', publisher: 'TollBit docs', href: 'https://docs.tollbit.com/docs/verifying-a-property', checked: '2026-11-14' },
+ { title: 'What is Pay Per Crawl?', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/ai-crawl-control/features/pay-per-crawl/what-is-pay-per-crawl/', checked: '2026-11-15' },
+ { title: 'Pay Per Use: when AI uses your work, you should get paid', publisher: 'Cloudflare blog', href: 'https://blog.cloudflare.com/pay-per-use/', checked: '2026-11-14' },
+ { title: 'Making AI search smarter', publisher: 'Cloudflare blog', href: 'https://blog.cloudflare.com/making-ai-search-smarter/', checked: '2026-11-15' },
+ { title: 'Analyze AI traffic', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/ai-crawl-control/features/analyze-ai-traffic/', checked: '2026-11-15' },
+ { title: 'Set a pay per crawl price', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/ai-crawl-control/features/pay-per-crawl/use-pay-per-crawl-as-site-owner/set-a-pay-per-crawl-price/', checked: '2026-11-15' },
+ { title: 'Enable in account settings', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/ai-crawl-control/features/pay-per-crawl/use-pay-per-crawl-as-site-owner/enable-in-account-settings/', checked: '2026-11-15' },
+ { title: 'AI Crawl Control with WAF', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/ai-crawl-control/configuration/ai-crawl-control-with-waf/', checked: '2026-11-15' },
+ { title: 'Manage payouts', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/ai-crawl-control/features/pay-per-crawl/use-pay-per-crawl-as-site-owner/manage-payouts/', checked: '2026-11-15' },
+ { title: 'Communicating license terms to AI agents', publisher: 'AWS WAF Developer Guide', href: 'https://docs.aws.amazon.com/waf/latest/developerguide/waf-ai-traffic-monetization-license-terms.html', checked: '2026-11-14' },
+ { title: 'Getting started with AI traffic monetization', publisher: 'AWS WAF Developer Guide', href: 'https://docs.aws.amazon.com/waf/latest/developerguide/waf-ai-traffic-monetization-getting-started.html', checked: '2026-11-14' },
+ { title: 'AWS WAF Bot Control rule group', publisher: 'AWS WAF Developer Guide', href: 'https://docs.aws.amazon.com/waf/latest/developerguide/aws-managed-rule-groups-bot.html', checked: '2026-11-14' },
+ { title: 'How to control and monetize AI bot traffic using Fastly and TollBit', publisher: 'Fastly', href: 'https://www.fastly.com/blog/how-to-control-and-monetize-ai-bot-traffic-using-fastly-and-tollbit', checked: '2026-11-14' },
+ { title: 'How Skyfire and Akamai enabled 2.5 million agentic transactions', publisher: 'Akamai', href: 'https://www.akamai.com/blog/security/how-skyfire-akamai-enabled-25-million-agentic-transactions', checked: '2026-11-14' },
+ { title: 'Product', publisher: 'MonetizationOS', href: 'https://monetizationos.com/product', checked: '2026-11-14' },
+ { title: 'MonetizationOS demonstrates machine payments at Stripe Sessions 2026', publisher: 'MonetizationOS blog', href: 'https://monetizationos.com/blog/monetizationos-demonstrates-machine-payments-at-stripe-sessions-2026', checked: '2026-11-14' },
+ { title: 'MonetizationOS documentation', publisher: 'MonetizationOS', href: 'https://docs.monetizationos.com/', checked: '2026-11-14' },
+ { title: 'Control AI crawlers and assistants separately in Vercel Firewall', publisher: 'Vercel changelog', href: 'https://vercel.com/changelog/block-ai-crawlers-and-assistants-independently-with-managed-rules', checked: '2026-11-15' },
+ { title: 'Block AI crawlers and bots', publisher: 'Netlify Docs', href: 'https://docs.netlify.com/build/build-with-ai/block-ai-crawlers/', checked: '2026-11-15' },
+ { title: 'Bot management overview', publisher: 'Google Cloud Armor documentation', href: 'https://docs.cloud.google.com/armor/docs/bot-management', checked: '2026-11-15' },
+ { title: 'Configure bot protection for Web Application Firewall with Azure Front Door', publisher: 'Microsoft Learn', href: 'https://learn.microsoft.com/en-us/azure/web-application-firewall/afds/waf-front-door-policy-configure-bot-protection', checked: '2026-11-15' },
+ { title: 'AWS (S3, CloudFront, ALB)', publisher: 'TollBit docs', href: 'https://docs.tollbit.com/docs/aws-s3-cloudfront-alb', checked: '2026-11-14' },
+ { title: 'Select crawlers to charge', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/ai-crawl-control/features/pay-per-crawl/use-pay-per-crawl-as-site-owner/select-crawlers-to-charge/', checked: '2026-11-15' },
+ { title: 'TollBit homepage', publisher: 'TollBit', href: 'https://tollbit.com/', checked: '2026-11-14' },
+ { title: 'Web Bot Auth', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/', checked: '2026-11-15' },
+ { title: 'Redefine trust with Web Bot Authentication', publisher: 'Akamai', href: 'https://www.akamai.com/blog/security/redefine-trust-web-bot-authentication', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'Agent-payment rails',
+ items: [
+ { title: 'KYAPay Token (draft-skyfire-oauth-kyapay-token)', publisher: 'IETF Datatracker', href: 'https://datatracker.ietf.org/doc/draft-skyfire-oauth-kyapay-token/', checked: '2026-11-14' },
+ { title: 'Payments and settlement', publisher: 'Skyfire docs', href: 'https://docs.skyfire.xyz/docs/payments-settlement', checked: '2026-11-14' },
+ { title: 'Service directory (public API)', publisher: 'Skyfire', href: 'https://api.skyfire.xyz/api/v1/directory/services', checked: '2026-11-14' },
+ { title: 'The "Payment" HTTP Authentication Scheme (draft-httpauth-payment)', publisher: 'IETF Datatracker', href: 'https://datatracker.ietf.org/doc/draft-httpauth-payment/', checked: '2026-11-15' },
+ { title: 'x402 overview', publisher: 'Coinbase Developer Platform', href: 'https://docs.cdp.coinbase.com/x402/welcome', checked: '2026-11-14' },
+ { title: 'Deploy at the edge', publisher: 'Supertab Connect docs', href: 'https://connect-docs.supertab.co/guides/deploy-cdn', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'Retrieval marketplaces',
+ items: [
+ { title: 'Dappier data_model_id list: publisher search models and prices', publisher: 'Dappier', href: 'https://agents.dappier.com/zcj/cv3gcduzbjkm/guides/publisher-data-models.md', checked: '2026-11-15' },
+ { title: 'Use cases', publisher: 'Cashmere', href: 'https://cashmere.io/use-cases', checked: '2026-11-14' },
+ { title: 'Cashmere.io wants better AI licensing deals for publishers', publisher: 'Publishers Weekly', href: 'https://www.publishersweekly.com/pw/by-topic/digital/content-and-e-books/article/99705-the-stealthy-startup-promising-better-ai-publishing-licensing-deals.html', checked: '2026-11-14' },
+ { title: 'Creating and configuring collections', publisher: 'Cashmere support', href: 'https://cashmere.io/support/creating-and-configuring-collections', checked: '2026-11-14' },
+ { title: 'Full-stack AI distribution', publisher: 'Valyu', href: 'https://www.valyu.ai/rev-share-partner-programme', checked: '2026-11-14' },
+ { title: 'Wiley Health & Life Sciences', publisher: 'Valyu docs', href: 'https://docs.valyu.ai/guides/wiley-hls', checked: '2026-11-14' },
+ { title: 'AskNews partners with top publishers to serve booming news analytics market', publisher: 'AskNews', href: 'https://asknews.app/en/blog-articles/asknews-partners-with-top-publishers-to-serve-booming-news-analytics-market', checked: '2026-11-14' },
+ { title: 'Publishers quietly cut \'six-figure\' deals via Snowflake\'s AI licensing platform', publisher: 'Digiday', href: 'https://digiday.com/media/publishers-quietly-cut-six-figure-deals-via-snowflakes-ai-licensing-platform/', checked: '2026-11-14' },
+ { title: 'Cortex Knowledge Extensions', publisher: 'Snowflake documentation', href: 'https://docs.snowflake.com/en/user-guide/snowflake-cortex/cortex-knowledge-extensions/cke-overview', checked: '2026-11-15' },
+ { title: 'Dow Jones Factiva surpasses 8,000 licensed sources for GenAI use', publisher: 'Dow Jones', href: 'https://www.dowjones.com/press-room/dow-jones-factiva-surpasses-8000-licensed-sources-for-genai-use/', checked: '2026-11-14' },
+ { title: 'Dow Jones Investor Briefing, March 16, 2026 (News Corp Form 8-K, Exhibit 99.1)', publisher: 'U.S. SEC, EDGAR', href: 'https://www.sec.gov/Archives/edgar/data/1564708/000156470826000057/dowjones2026investorbrie.htm', checked: '2026-11-14' },
+ { title: 'Discover services (Bazaar)', publisher: 'Coinbase Developer Platform', href: 'https://docs.cdp.coinbase.com/x402/bazaar', checked: '2026-11-15' },
+ { title: 'For publishers', publisher: 'xpay documentation', href: 'https://xpay.sh/docs/publishers', checked: '2026-11-15' },
+ { title: 'Model Context Protocol specification, revision 2026-07-28', publisher: 'Model Context Protocol', href: 'https://modelcontextprotocol.io/specification/2026-07-28', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Answer engines',
+ items: [
+ { title: 'Gist Ads discontinued', publisher: 'ProRata.ai', href: 'https://gist.ai/gist-ads-discontinued', checked: '2026-11-14' },
+ { title: 'ProRata Online Master Services Agreement (MSA)', publisher: 'ProRata.ai', href: 'https://gist.ai/support/ga/wpnewpartner', checked: '2026-11-15' },
+ { title: 'Gist Answers for WordPress: user guide', publisher: 'ProRata.ai', href: 'https://gist.ai/support/ga/wpuserguide', checked: '2026-11-14' },
+ { title: 'How Perplexity\'s new revenue model works, according to its head of publisher partnerships', publisher: 'Digiday', href: 'https://digiday.com/media/how-perplexity-new-revenue-model-works-according-to-its-head-of-publisher-partnerships/', checked: '2026-11-14' },
+ { title: 'Introducing Comet Plus', publisher: 'Perplexity', href: 'https://www.perplexity.ai/hub/blog/introducing-comet-plus', checked: '2026-11-15' },
+ { title: 'Perplexity Crawlers', publisher: 'Perplexity docs', href: 'https://docs.perplexity.ai/docs/resources/perplexity-crawlers', checked: '2026-11-15' },
+ { title: 'Index by Parallel', publisher: 'Parallel', href: 'https://index.parallel.ai/', checked: '2026-11-14' },
+ { title: 'Parallel Web Systems bots', publisher: 'Parallel', href: 'https://parallel.ai/parallel-web-systems-bots', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'Training data',
+ items: [
+ { title: 'AI rights and the future of publishing with Created by Humans founder Trip Adler', publisher: 'Written Word Media', href: 'https://www.writtenwordmedia.com/ai-rights-and-the-future-of-publishing-with-created-by-humans-founder-trip-adler/', checked: '2026-11-14' },
+ { title: 'Author Overview', publisher: 'Created by Humans help center', href: 'https://support.createdbyhumans.ai/hc/en-us/articles/30891024362523-Author-Overview', checked: '2026-11-14' },
+ { title: 'Video', publisher: 'Protege', href: 'https://withprotege.ai/data-provider/video', checked: '2026-11-14' },
+ { title: 'How AI content licensing deals work: structures, terms, and what gets negotiated', publisher: 'Troveo', href: 'https://www.troveo.ai/resources/ai-data-licensing-deals', checked: '2026-11-14' },
+ { title: 'Turn your video library into recurring revenue', publisher: 'Troveo', href: 'https://www.troveo.ai/content-owners', checked: '2026-11-14' },
+ { title: 'llms.txt', publisher: 'Wirestock', href: 'https://wirestock.io/llms.txt', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'Court settlements',
+ items: [
+ { title: 'Order granting final approval of class action settlement (Dkt. 680), Bartz v. Anthropic', publisher: 'U.S. District Court, Northern District of California', href: 'https://assets-us-01.kc-usercontent.com:443/1eeb16db-4934-006e-40a6-38fa91285ebb/36cce252-a32c-4cbb-a624-31dd8ac9574c/2026-07-20%20Order%20Granting%20Final%20_dckt%20680_0_.pdf', checked: '2026-11-15' },
+ { title: 'Bartz v. Anthropic settlement website', publisher: 'Settlement administrator', href: 'https://www.anthropiccopyrightsettlement.com/', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Changed, closed or announced',
+ items: [
+ { title: 'AI visibility measurement for publishers and brands', publisher: 'ScalePost', href: 'https://www.scalepost.ai/', checked: '2026-11-14' },
+ { title: 'Human Native is joining Cloudflare', publisher: 'Cloudflare blog', href: 'https://blog.cloudflare.com/human-native-joins-cloudflare/', checked: '2026-11-14' },
+ { title: 'Human Native is joining Cloudflare', publisher: 'Human Native', href: 'https://www.humannative.ai/', checked: '2026-11-14' },
+ { title: 'Human Native Ltd (15609120): persons with significant control', publisher: 'Companies House', href: 'https://find-and-update.company-information.service.gov.uk/company/15609120/persons-with-significant-control', checked: '2026-11-14' },
+ { title: 'Next Net home page', publisher: 'Next Net', href: 'https://www.nextnet.ai', checked: '2026-11-15' },
+ { title: 'Amazon may launch a marketplace where media sites can sell their content to AI companies', publisher: 'TechCrunch', href: 'https://techcrunch.com/2026/02/10/amazon-may-launch-a-marketplace-where-media-sites-can-sell-their-content-to-ai-companies/', checked: '2026-11-13' },
+ ],
+ },
+ {
+ name: 'Training and content signals',
+ items: [
+ { title: 'HTTP Message Signatures for automated traffic Architecture (draft-meunier-web-bot-auth-architecture)', publisher: 'IETF, individual Internet-Draft', href: 'https://datatracker.ietf.org/doc/draft-meunier-web-bot-auth-architecture/', checked: '2026-11-14' },
+ { title: 'Giving users choice with Cloudflare\'s new Content Signals Policy', publisher: 'Cloudflare blog', href: 'https://blog.cloudflare.com/content-signals-policy/', checked: '2026-11-14' },
+ { title: 'robots.txt controls', publisher: 'Anthropic', where: 'Anthropic help center, article 8896518 (no link)', checked: '2026-11-15' },
+ { title: 'Helping websites express AI training preferences', publisher: 'OpenAI', href: 'https://openai.com/index/helping-websites-express-ai-training-preferences/', checked: '2026-11-14' },
+ { title: 'Google\'s common crawlers', publisher: 'Google for Developers', href: 'https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers', checked: '2026-11-15' },
+ { title: 'Your content, your choice: clearer preferences for training and grounding', publisher: 'Microsoft, Bing blog', href: 'https://blogs.bing.com/search/2026/10/your-content-your-choice-clearer-preferences-for-training-and-grounding/', checked: '2026-11-15' },
+ ],
+ },
+];
+const FLAT: { href?: string }[] = SOURCES.flatMap((g) => g.items);
+/** Entry N's URL, for an inline link to the same source. */
+const src = (n: number, anchor = '') => `${FLAT[n - 1].href}${anchor}`;
+
+// What FORA takes: FORA's cells of the GlanceCard, under the shared row labels.
+const FORA_GLANCE: Record = {
+ payer: FORA_CELLS.payer,
+ fees: FORA_CELLS.fees,
+ reporting: `Each sale, and daily income as a CSV, in your operator's publisher console. No bot-traffic report ${cite(1)}.`,
+ contract: `An agreement with your Exchange operator. With each purchase the buyer signs an acceptance of your terms, which the console can check ${cite(1, 2)}.`,
+ status: `${statusTag('live')} ${SEE_ABOUT}`,
+ dns: `No record to add, and nothing delegated ${cite(1)}.`,
+ cdn: FORA_CELLS.cdn,
+ firewall: `Your rules run first. They must let signed requests and your FORA file through ${cite(1, 3)}.`,
+ site: `One file, fora.json, and a sitemap the operator imports your pages from. No keys to make, no catalog to push ${cite(1)}.`,
+ search: `Google's, Bing's and Apple's crawlers pass. The rest is your choice: see ${a('#together', 'Search traffic')} ${cite(2, 3)}.`,
+ upkeep: `Keep your terms current with your Exchange. A change to your FORA file needs no redeploy ${cite(1)}.`,
+ leaving: FORA_CELLS.leaving,
+};
+const GLANCE_TITLES = ['FORA for your revenue team', 'What FORA takes on your stack'];
+const GLANCE = GLANCE_GROUPS.map((g, i) => ({
+ title: GLANCE_TITLES[i],
+ group: g.name,
+ rows: g.rows.map((r) => ({ label: r.label, cell: FORA_GLANCE[r.key] })),
+}));
+
+const SHORT_ANSWER = [
+ { title: 'Seven kinds of channel sell future use, and court settlements pay for the past.', text: 'Some charge bots at your CDN, and some hold a copy of your content and sell search over it. Others are the AI company itself, paying a share, a fee or by contribution.' },
+ { title: 'They differ on who sets the price.', text: 'At a bot paywall you set it per page. In an answer engine\'s pool or Google\'s pilots, the buyer decides your share. In Microsoft\'s marketplace the terms aren\'t public.' },
+ { title: 'FORA lets anyone run a channel you price yourself.', text: 'It is an open protocol and free, open-source software, so any company can run a FORA marketplace, an Exchange. You sell through one, several or your own, and set terms and a price for each use of your pages. The operator bills the buyer and pays you. It runs beside your other deals.' },
+];
+
+const KEY_FACTS = [
+ { value: '400+', text: `publications Google pays in its News AI pilot ${cite(9)}` },
+ { value: '$203.0 million', text: `in aggregate AI licensing deals, in Reddit's S-1 filing ${cite(10)}` },
+ { value: 'NOK 45 million', text: `a year to Kopinor from the National Library of Norway ${cite(20)}` },
+ { value: 'Nearly 20%', text: `of nearly 7,000 TollBit sites have earned, by TollBit's CEO in April 2026, unaudited ${cite(25)}` },
+];
+
+// The map: FORA's row first, then one group per category of channel.
+const MAP_COLUMNS = ['Player', 'What\'s sold', 'Who pays today', 'Fee, payout and reporting', 'Who sets the price', 'Unit', 'What your engineers change', 'How FORA works with it'];
+const MAP_FORA = {
+ player: 'FORA',
+ cells: [
+ `A license to each page or file, under offers your Exchange signs ${cite(4)}`,
+ '—',
+ `FORA itself is free. Each sale in your operator's console ${cite(1)}. The console's fee and payout: see ${ABOUT}`,
+ `You, per term, by use, country and buyer type ${cite(5)}`,
+ `Free, flat, or per unit you name: text tokens (about ¾ of a word), pages, seconds or your own ${cite(4, 5)}`,
+ `An edge function and one file; Cloudflare, Fastly or CloudFront only ${cite(1, 2)}. See ${a('#fora', 'What FORA takes')}`,
+ '—',
+ ],
+};
+const MAP = [
+ {
+ name: '1. Bot paywalls at your CDN',
+ rows: [
+ { player: 'TollBit', cells: [
+ `A license to one page: summary, full display or custom; no training ${cite(26, 27)}`,
+ `Few buyers named, none confirmed as paying ${cite(28, 29)}. TNL Mediagene reports "initial revenue from AI buyers"; TollBit's CEO says nearly 20% of nearly 7,000 sites have earned (April 2026, unaudited) ${cite(25, 30)}`,
+ `The buyer pays a fee on top of your rate; its size isn't published ${cite(27)}. Shows bot traffic ${cite(31)}`,
+ 'You',
+ `Per page, rate set per 1,000 ${cite(27)}`,
+ `DNS records, a CDN rule and firewall allow rules; all your logs shared; the bot list kept by hand ${cite(32, 33)}`,
+ `Runs beside FORA's edge in a CDN rule order you write ${cite(32)}. Its rates shown to FORA agents: ${statusTag('design')}`,
+ ] },
+ { player: 'Cloudflare', cells: [
+ `Pay Per Crawl ${statusTag('closed-beta')}: each successful crawl ${cite(34)}. Pay Per Use ${statusTag('beta')}: a use the buyer defines ${cite(35)}`,
+ `Pay Per Use: Ceramic.ai and You.com in pilots ${cite(35, 36)}. No Pay Per Crawl buyer named; no amounts published`,
+ `Fee not published. Shows bot traffic ${cite(37)}`,
+ `Pay Per Crawl: you ${cite(38)}. Pay Per Use: the buyer ${cite(35)}`,
+ 'Per crawl; per reported use',
+ `Cloudflare's nameservers or Cloudflare-managed DNS ${cite(39)}, beta admission ${cite(34)}, security rules reconciled ${cite(40)}, a new Stripe Connect account ${cite(41)}`,
+ `FORA's edge runs as a Worker after Cloudflare's rules ${cite(1, 3)}`,
+ ] },
+ { player: 'AWS WAF', cells: [
+ `Access to requests a rule matches; no license ${cite(42)}`,
+ 'No buyer or publisher named',
+ `To a USDC wallet or a Stripe account; AWS lists each settlement ${cite(43)}`,
+ 'You',
+ `Per request, in USDC ${cite(43)}`,
+ `CloudFront, firewall rules per price, a wallet or Stripe account, your own log pipeline ${cite(43, 44)}`,
+ `FORA's edge runs on CloudFront after AWS WAF ${cite(1, 3)}`,
+ ] },
+ { player: 'Fastly, Akamai', cells: [
+ `Bot detection; the sale goes to TollBit, Skyfire or Supertab ${cite(45)}`,
+ `Through partners. Akamai reports 2.5 million Skyfire transactions by one unnamed data provider ${cite(46)}`,
+ 'The partner\'s',
+ 'The partner',
+ 'The partner\'s',
+ 'Bot management plus the partner\'s rule',
+ `FORA's edge runs on Fastly, not on Akamai ${cite(1, 2)}`,
+ ] },
+ { player: 'MonetizationOS', cells: [
+ `Access plans for readers and bots ${cite(47)}`,
+ 'No public case of a bot paying',
+ '—',
+ 'You',
+ 'A plan or meter',
+ 'A proxy on your CDN',
+ 'Runs in series with FORA\'s edge',
+ ] },
+ ],
+ },
+ {
+ name: '2. Agent-payment rails',
+ rows: [
+ { player: 'Skyfire (KYAPay)', cells: [
+ `Signed identity and payment passes, charged after delivery ${cite(59, 60)}`,
+ `No AI company named. 81 services listed, one news site at $0.01 a use ${cite(61)}`,
+ `Settles in up to about 51 hours ${cite(60)}`,
+ 'The seller',
+ 'The seller\'s price per service',
+ 'Register a seller service; check and charge passes',
+ 'Could fund an agent\'s account at a FORA Exchange; FORA ships no adapter',
+ ] },
+ { player: 'x402, MPP', cells: [
+ `Payment for one request ${cite(62, 63)}`,
+ 'No AI company named as a buyer of publisher content',
+ 'Settles per request',
+ 'The seller',
+ 'Per request',
+ 'A payment handler or gateway at your edge',
+ 'Could carry the payment for a FORA sale; FORA ships no adapter',
+ ] },
+ { player: 'Supertab Connect', cells: [
+ 'RSL licenses through a license server',
+ 'No licensee named',
+ '—',
+ 'You',
+ 'Per crawl, call, document, text token and more',
+ `A CDN SDK or WordPress plugin, and a License: line ${cite(64)}`,
+ 'Its RSL file can sit beside FORA\'s terms',
+ ] },
+ ],
+ },
+ {
+ name: '3. Retrieval marketplaces',
+ rows: [
+ { player: 'Dappier', cells: [
+ `Search over your corpus as a "data model" ${cite(65)}`,
+ 'Agents prepay credit through ZeroClick. No AI company named',
+ '—',
+ 'You',
+ `Per query, $0.003 to $0.02 ${cite(65)}`,
+ 'An RSS feed or a script tag',
+ 'Runs beside FORA, from its own copy of your content',
+ ] },
+ { player: 'Cashmere', cells: [
+ 'Search results over your catalog, which it hosts',
+ `Perplexity, named by Cashmere and press ${cite(66, 67)}. No payment confirmed by a buyer or publisher`,
+ '—',
+ `You, per collection ${cite(68)}`,
+ 'Per million text tokens',
+ 'Upload titles by API, S3 or SFTP',
+ 'Runs beside FORA, from its own copy of your content',
+ ] },
+ { player: 'Valyu', cells: [
+ 'Cited passages from your archive',
+ 'Buyers prepay',
+ `Payouts promised "monthly above a threshold"; none shown ${cite(69)}`,
+ `Valyu: Wiley's health and life sciences journals at $12 per 1,000 results ${cite(70)}`,
+ 'Per result',
+ 'Hand over files',
+ 'Runs beside FORA, from its own copy of your archive',
+ ] },
+ { player: 'AskNews', cells: [
+ 'Search over "synthetic" versions of your articles',
+ 'Buyers on published plans. No publisher confirms a payout',
+ `Said in 2025 that 50% of its revenue goes to publishers ${cite(71)}`,
+ 'AskNews',
+ 'Credits and text tokens',
+ 'It crawls you or reads your feed; its crawler\'s user agent and IP ranges aren\'t published',
+ 'Runs beside FORA, from its own copy of your content',
+ ] },
+ { player: 'Snowflake (CKE)', cells: [
+ 'A search service over your text, inside Snowflake',
+ `Buyers pay through Snowflake, or on terms struck with you directly. "six-figure deals" with financial institutions (Snowflake, via Digiday) ${cite(72)}`,
+ `By subscription; no usage-based billing ${cite(73)}`,
+ 'You',
+ 'Subscription',
+ 'Build the service in a Snowflake account',
+ 'Runs beside FORA: CKE sells inside Snowflake, FORA at your site',
+ ] },
+ { player: 'Dow Jones Factiva', cells: [
+ `Your articles, relicensed to companies for AI use ${cite(74)}`,
+ `"~20 GenAI deals" with companies, none named (Dow Jones) ${cite(75)}`,
+ 'Royalties to publishers by agreement',
+ 'Dow Jones',
+ 'A subscription per project, metered in text tokens',
+ 'Sign an agreement',
+ 'Runs beside FORA, from its own copy of your articles',
+ ] },
+ { player: 'Newstex, xpay', cells: [
+ `Newstex: "one full licensed story body" per call ${cite(76)}. xpay: search over your feed ${cite(77)}`,
+ `Newstex's story route: 3 calls by 1 payer in 30 days, by x402 Bazaar's count ${cite(76)}. xpay names no buyer`,
+ `xpay keeps 5% ${cite(77)}`,
+ 'Newstex; at xpay, you',
+ `Per call: Newstex $0.01 a search, $0.10 a story; xpay $0.01 to $1.00 a query ${cite(76, 77)}`,
+ `xpay: paste an RSS feed. "Your site, CDN, and DNS stay completely untouched", so nothing is enforced at your site ${cite(77)}`,
+ 'Runs beside FORA, from its own copy of your feed',
+ ] },
+ ],
+ },
+ {
+ name: '4. Answer engines that pay a share',
+ rows: [
+ { player: 'ProRata, Gist', cells: [
+ 'Use of your content in Gist answers',
+ `ProRata itself; no payout published. Its Gist Ads network ended 2026-09-01; Gist Answers still runs ${cite(79)}`,
+ `Monthly above $1,000, with monthly statements per article and an audit right ${cite(80)}`,
+ 'ProRata',
+ `A share of 50% of net revenue, split per answer by attribution ${cite(80)}`,
+ `A script tag or WordPress plugin; your feed open to its crawler, whose user agent and IP ranges aren't published ${cite(81)}`,
+ `Runs beside FORA on one site; its MSA is non-exclusive ${cite(80)}`,
+ ] },
+ { player: 'Perplexity', cells: [
+ 'Use in Perplexity\'s products',
+ `Perplexity itself. Of Comet Plus, its head of publisher partnerships said "Real cash payments starting immediately" (Digiday, 2025) ${cite(82)}. Comet Plus is on none of Perplexity's pricing pages`,
+ '—',
+ 'Perplexity',
+ `Comet Plus: a subscription pool by visits, citations and agent actions ${cite(83)}`,
+ `A contract; its two crawlers allowed by user agent and published IP ranges. Nothing metered at your site ${cite(84)}`,
+ `Runs beside FORA on one site; you choose what each of its crawlers gets at FORA's edge ${cite(3)}`,
+ ] },
+ { player: 'Index by Parallel', cells: [
+ 'Use in Parallel\'s agents\' work',
+ `Its home page marks monetization "Coming Soon" ${cite(85)}`,
+ 'Terms private',
+ 'Parallel',
+ 'Each source\'s estimated contribution to the answer',
+ 'A partnership agreement',
+ `FORA's edge refuses its crawler on licensed paths unless you allow it ${cite(3)}`,
+ ] },
+ ],
+ },
+ {
+ name: '5. AI companies buying directly',
+ rows: [
+ { player: 'Google', cells: [
+ `News AI pilot: "enhanced content" for AI features ${cite(9)}. AI contribution pilot: content that "contributes significantly" to an AI answer ${cite(11)}. In the EU, previews (ENP) ${cite(12)}`,
+ `Google itself: "more than 400 publications" in the News AI pilot (Google); about 100 publishers in the contribution pilot (The Information, via TNW) ${cite(9, 13)}`,
+ `Contribution pilot: Google decides the amount and shows one monthly number in Search Console ${cite(11)}`,
+ 'Google',
+ 'Not public',
+ `A verified Search Console property, the program terms, payment details; Googlebot kept unblocked and unmetered ${cite(12)}`,
+ `FORA's edge passes Googlebot free; Google's AI uses stay under Google's own controls ${cite(3)}`,
+ ] },
+ { player: 'Microsoft PCM', cells: [
+ 'Grounding and display in Copilot and partners\' products',
+ `Copilot; Microsoft is "beginning to onboard" Yahoo. No payment confirmed by a primary source ${cite(14)}`,
+ 'Pricing, unit and fee not public',
+ 'Not public; pilot contracts negotiated one by one',
+ 'Not public',
+ 'A contract; nothing documented on your site',
+ 'Runs beside FORA on one site, if your PCM contract allows it',
+ ] },
+ { player: 'Direct deals', cells: [
+ 'Display, archive access, often training',
+ `Named buyers: OpenAI, Google, Meta, Amazon, Microsoft, Mistral, Anthropic ${cite(15, 16)}. Reddit filed $203.0 million in aggregate deals ${cite(10)}`,
+ 'Private contracts',
+ 'Negotiated',
+ 'Mostly a flat annual fee',
+ 'A contract; a feed or a crawl',
+ `The buyer can buy through FORA under a deal term ${cite(5)}`,
+ ] },
+ ],
+ },
+ {
+ name: '6. Training data',
+ rows: [
+ { player: 'Created by Humans', cells: [
+ 'AI rights to books, bundled per deal',
+ 'No deal confirmed',
+ `Its CEO: "we take a 20 % fee" ${cite(87)}`,
+ `Negotiated by CbH; you can opt out within 7 days ${cite(88)}`,
+ 'Per deal',
+ 'Claim your books by ISBN',
+ 'Runs beside FORA, from its own copy of your books',
+ ] },
+ { player: 'Defined.ai, Protege, Wirestock, Troveo', cells: [
+ 'Datasets for training',
+ `By their own figures: Protege "Tens of Millions", Wirestock "over $15 million" paid; Troveo gives two different totals ${cite(89, 90, 91, 92)}`,
+ '—',
+ 'Negotiated',
+ 'Per deal',
+ 'Hand over your archive',
+ `Runs beside FORA, from its own copy. Where a vendor holds an exclusive training grant, you can show a FORA training term only to that buyer ${cite(5)}`,
+ ] },
+ ],
+ },
+ {
+ name: '7. Collectives and rights agencies',
+ rows: [
+ { player: 'ILC, CLA, DPCMO, Kopinor', cells: [
+ 'A collective license across members',
+ `ILC: Yahoo Scout, its first licensee ${cite(21)}. Kopinor: NOK 45 million a year from the National Library of Norway ${cite(20)}`,
+ 'Split by each scheme\'s rule',
+ 'The collective',
+ 'A pool',
+ 'Join and accept the terms',
+ `A pool on FORA: see ${a('/resources/transparent-collective/', 'A fair licensing collective')}`,
+ ] },
+ { player: 'CCC', cells: [
+ `Rights only, no copy: annual licenses (a pool) and AI Summarization, one summary per order ${cite(22, 23)}`,
+ `Buyers pay CCC; no AI payment is public. One of nine titles probed sold AI Summarization, at 19.10 USD a summary ${cite(24)}`,
+ '—',
+ 'CCC for annual licenses; you, per order',
+ 'Per organization a year; per summary',
+ 'An agreement, a title list and a fee table in CCC\'s portal; nothing at your site',
+ 'A CCC order maps to a FORA term; CCC documents no buyer API',
+ ] },
+ ],
+ },
+ {
+ name: '8. Court settlements (past use)',
+ rows: [
+ { player: 'Bartz v. Anthropic', cells: [
+ `Release of claims for past copying of 440,490 claimed books ${cite(93)}`,
+ `Anthropic, into a $1.5 billion fund the court approved ${cite(93)}`,
+ `About $3,000 a work, by the court's estimate; first payments "on or before November 15, 2026" ${cite(93, 94)}`,
+ 'The court',
+ 'Per work',
+ 'A claim through the administrator',
+ 'Neither: a one-time remedy, not a license',
+ ] },
+ ],
+ },
+ {
+ name: 'Changed or closed',
+ rows: [
+ { player: 'ScalePost', cells: [
+ `Was a licensing marketplace; now measures AI visibility ${cite(95)}`,
+ 'No content sale shown',
+ '—', '—', '—',
+ 'CDN logs',
+ 'Reads logs; runs beside FORA\'s edge',
+ ] },
+ { player: 'Human Native', cells: [
+ `Was a training-data marketplace ${cite(96)}`,
+ `Closed: its site is now one letter, "Human Native is joining Cloudflare" ${cite(97)}. Cloudflare holds 75% or more ${cite(98)}`,
+ '—', '—', '—', '—', '—',
+ ] },
+ ],
+ },
+];
+
+const RELATIONS = [
+ { title: 'Sell the same page.', text: 'Bot paywalls, retrieval marketplaces, AI companies buying directly, and Supertab Connect\'s license server each sell what a FORA Exchange sells. Each sale goes to one of them, and they run side by side on one site.' },
+ { title: 'Work alongside FORA.', text: 'The payment rails, at settlement; answer engines; and training-data deals.' },
+ { title: 'Could run on FORA.', text: 'A retrieval marketplace, a collective or a rights agency can run a FORA Exchange and Broker over its own catalog, keep its search, and reach every FORA agent.' },
+];
+
+const ENGINEERS_FULL = [
+ { title: 'TollBit and Cloudflare:', text: `the setup is on their deep pages, ${a('/resources/content-marketplaces/tollbit/', 'TollBit')} and ${a('/resources/content-marketplaces/cloudflare/', 'Cloudflare')}.` },
+ { title: 'AWS WAF:', text: 'the site on CloudFront. A global web ACL with Bot Control, its blocking rules set to Count so bots reach a price. One Monetize rule per price, on AWS\'s bot labels. A USDC wallet or Stripe account, and your own log pipeline.' },
+ { title: 'Fastly, Akamai:', text: 'for TollBit, a backend and VCL snippet (Fastly) or a Conditional Origin and Cloudlet (Akamai).' },
+ { title: 'Google:', text: 'for ENP, a DNS TXT record to verify the property. No CDN rule, feed or log export.' },
+ { title: 'Microsoft PCM:', text: 'no ingestion format, DNS or CDN change, bot rule or log feed is documented.' },
+ { title: 'CCC:', text: 'a rightsholder agreement. No DNS, CDN or log change, and nothing enforced at your site.' },
+ { title: 'FORA:', text: `see ${a('#together', 'What changes on your CDN')}.` },
+];
+
+// hub-4: how money reaches a publisher, one tile per channel and FORA's.
+const MONEY_TILES = [
+ { title: '1 · Bot paywalls', flow: ['AI bot', 'TollBit, Cloudflare', 'You'], lines: ['Pays per page or crawl, at your CDN', 'Price: yours. TollBit adds a buyer fee', 'whose size it doesn\'t publish'] },
+ { title: '2 · Payment rails', flow: ['Agent', 'x402, MPP', 'You'], lines: ['One payment for one request', 'No license travels with it', 'Skyfire: charged after delivery'] },
+ { title: '3 · Retrieval marketplaces', flow: ['Buyer', 'Marketplace, your copy', 'You'], lines: ['Pays per query or result', 'Price: yours at Dappier and Cashmere,', 'the marketplace\'s at Valyu and AskNews'] },
+ { title: '4 · Answer engines', flow: ['Their revenue', 'ProRata, Perplexity', 'You'], lines: ['Pays a share, from a pool', 'ProRata: 50% of net revenue', 'Split by the use they count'] },
+ { title: '5 · Buying directly', flow: ['Google, Microsoft, AI companies', 'You'], lines: ['Pays from its own programs or deals', 'Google decides the amount', 'Direct deals: mostly a flat annual fee'] },
+ { title: '6 · Training data', flow: ['AI lab', 'Dataset vendor', 'You'], lines: ['Pays per deal, for bulk copies', 'Read whether the grant is exclusive', 'Created by Humans: a 20% fee'] },
+ { title: '7 · Collectives', flow: ['Licensee', 'Collective', 'Members'], lines: ['One license across many members', 'Split by the scheme\'s own rule', 'Kopinor: NOK 45 million a year'] },
+ { title: '8 · Court settlements', flow: ['Anthropic', '$1.5 billion fund', 'Claimants'], lines: ['Paid once, for past copying', 'About $3,000 a work, court\'s estimate', 'No future price, no license'] },
+ { title: 'FORA', flow: ['Agent', 'Exchange operator', 'You'], lines: ['Charges the agent\'s own account', 'The operator bills and pays you'], fora: true },
+];
+
+const CAUTIONS = [
+ { title: 'TollBit\'s fetcher must get through.', text: `If its requests "are blocked, challenged, or rate-limited, your Agent Site will not work correctly" (${a(src(33), 'Verifying a property')}).` },
+ { title: 'Bot pages can reach readers.', text: `On CloudFront, steps done out of order mean "Agent Site responses can be cached and then served to human visitors" (${a(src(54), 'CloudFront')}).` },
+ { title: 'Your firewall decides first.', text: `At Cloudflare, "WAF custom rules are enforced before pay per crawl", so a crawler one of them blocks is never offered a price (${a(src(40), 'WAF')}).` },
+ { title: 'Search can suffer.', text: `Charging a search crawler "may negatively impact your site's SEO performance" (${a(src(55), 'Select crawlers')}).` },
+];
+
+// The seven sections after bot paywalls: a heading and HTML paragraphs.
+const KINDS = [
+ {
+ title: 'Payment rails: one payment per request',
+ paras: [
+ `x402 and MPP carry a payment for one request. Skyfire's passes are charged after delivery, and settlement takes up to about 51 hours (${a(src(60), 'Skyfire')}). No rail carries a license. Supertab Connect adds an RSL license server and names no licensee.`,
+ ],
+ },
+ {
+ title: 'Retrieval marketplaces: they hold a copy and sell search over it',
+ paras: [
+ `You hand over your archive or feed, and the marketplace sells search results. At Dappier and Cashmere you set the price; at Valyu and AskNews the marketplace does and pays you a share. Snowflake's Cortex Knowledge Extensions sell access by subscription: "Usage-based billing with CKEs isn't supported" (${a(src(73), 'Snowflake')}). Most reach agents over MCP, which carries no price or license ${cite(78)}. Newstex and xpay charge per call over x402.`,
+ ],
+ },
+ {
+ title: 'Answer engines: the buyer measures and pays a share',
+ paras: [
+ `ProRata pays a pool of 50% of its net revenue, split per answer by how much came from you (${a(src(80), 'MSA')}). Perplexity's Comet Plus pools subscription revenue by "human visits, search citations, and agent actions" (${a(src(83), 'Perplexity')}).`,
+ `Deep pages: ${a('/resources/content-marketplaces/prorata-gist/', 'ProRata and Gist')}, ${a('/resources/content-marketplaces/perplexity/', 'Perplexity')}.`,
+ ],
+ },
+ {
+ title: 'AI companies buying directly',
+ paras: [
+ `In October 2026, Google said its News AI pilot pays "more than 400 publications worldwide" for "enhanced content" (${a(src(9), 'Google')}). Its AI contribution pilot pays when content "contributes significantly" to an answer in Gemini, AI Overviews and AI Mode, by Google's help text as Search Engine Roundtable published it ${cite(11)}. Google decides the amount and shows one monthly number in Search Console. Microsoft says PCM publishers "will be paid on delivered value", with pricing, unit and fee not public ${cite(14)}. Direct deals are private contracts, and some end a lawsuit: Folha's deal with OpenAI, signed with UOL in May 2026, ended the suit Folha filed in 2025 ${cite(17, 18)}.`,
+ `Deep page: ${a('/resources/content-marketplaces/microsoft-pcm/', 'Microsoft PCM')}.`,
+ ],
+ },
+ {
+ title: 'Training data: bulk copies under contract',
+ paras: [
+ `Defined.ai, Protege, Wirestock and Troveo package archives for model training, deal by deal, and Created by Humans does the same for books. Exclusivity is the term to read: Troveo says "around 95 percent" of its rights holders license exclusively (${a(src(90), 'Troveo')}).`,
+ ],
+ },
+ {
+ title: 'Collectives and rights agencies: one license, many works',
+ paras: [
+ `A collective licenses many members' content at once and splits the money by its own rule. ${a('/resources/transparent-collective/', 'A fair licensing collective')} compares eight schemes. CCC sells rights only: "This is a rights-only license and does not provide access to or authorize scraping of content" (${a(src(22), 'CCC')}). Its AI Summarization sells one summary per order, at the rightsholder's price.`,
+ ],
+ },
+ {
+ title: 'Court settlements: paid once, for the past',
+ paras: [
+ `In Bartz v. Anthropic, the court approved a $1.5 billion fund for 440,490 claimed books on July 20, 2026. Its "estimated per-work payment" is "approximately $3,000" (${a(src(93), 'final approval order')}). It sets no price for future use and grants no license.`,
+ ],
+ },
+];
+
+const AHEAD = [
+ { title: 'The buyer comes with the contract.', text: 'Google buys in its pilots and Copilot in Microsoft\'s marketplace; at ProRata and Perplexity, the channel itself is the buyer. On FORA, any agent can buy under your terms.' },
+ { title: 'One integration reaches many sites.', text: `A TollBit buyer uses one key, one card and one search across the network. TollBit's homepage says "over 10,000 sites", up from the "nearly 7,000" its CEO cited in April 2026 (${a(src(56), 'TollBit')}, vendor's figures). A FORA agent opens an account at each Exchange it buys from (${a(src(4), 'What is FORA?')}).` },
+ { title: 'Blocking bots on sites FORA isn\'t installed on.', text: `Cloudflare can refuse a crawler on any site it fronts, and says "More than 20% of the web sits behind Cloudflare’s network" (${a(src(36), 'Cloudflare')}). FORA's edge function refuses unlicensed AI bots on sites that install it, and nowhere else.` },
+ { title: 'Less to install.', text: `Google, CCC, Microsoft and Perplexity work through a contract, a portal or Search Console, with no paywall on your CDN. Google and Perplexity need their crawlers left unblocked. TollBit changes nothing on your origin or CMS. FORA needs its edge function on your CDN and its file on your site (see ${a('#fora', 'What FORA takes')}).` },
+ { title: 'Paying by how much your content shaped the answer.', text: `Google's contribution pilot pays when content "contributes significantly"; ProRata scores each source's share of an answer; Parallel says it pays by each source's estimated contribution ${cite(11, 80, 85)}. FORA cannot: the charge is fixed at the sale, and "a usage report never re-prices a purchase" (${a(src(5, '#metering-modes'), 'Licensing terms')}).` },
+ { title: 'Seeing which bots visit before you sell.', text: `TollBit and Cloudflare show which bots fetch what ${cite(31, 37)}. FORA Server records transactions and makes no bot-traffic report. It's open source, so an operator can add a log of what agents search for and what its Exchange offers them.` },
+ { title: 'A warranty in the license.', text: `TollBit's ${a(src(26), 'standard license')} has you warrant "all rights necessary to provide the Site Content" and indemnify the buyer. FORA cannot state either as a term. A term's obligations are all the buyer's, so a warranty fits only in a license document the term points to (${a(src(5, '#obligations'), 'Licensing terms')}).` },
+ { title: 'Payment and bot identity come from other standards.', text: `x402 and MPP settle per request, and Web Bot Auth was first drafted at Cloudflare ${cite(101)}. FORA defines no payment rail, by design, and its agents sign their requests with Web Bot Auth.` },
+];
+
+// The eight questions: "Others:" and "FORA:" on separate lines.
+const qa = (others: string, fora: string) => `Others: ${others}FORA: ${fora}`;
+const QUESTIONS = [
+ { title: 'How do you say what\'s for sale?', text: qa(
+ `most channels hold your terms in a dashboard or a contract. Supertab and ILC use RSL, and AWS points to an RSL file ${cite(21, 42, 64)}. Cloudflare's Content Signals and the __usage_ai-training group in robots.txt state a preference and no price ${cite(102)}. Anthropic, OpenAI, Google and Microsoft committed to honor that group ${cite(103, 104, 105, 106)}.`,
+ `your FORA file names the Exchanges that sell your content, and each holds your terms per page. FORA's ${a('/resources/ai-crawlers/', 'rules generator')} writes the training group. The ${a('/resources/standards-map/', 'standards map')} compares them all.`,
+ ) },
+ { title: 'How does a buyer find your content?', text: qa(
+ `each channel runs its own index under its own rules: TollBit's search, Cloudflare's list of charging domains, Dappier's and Valyu's search. Coinbase's x402 Bazaar ranks by relevance and by "recent call volume and unique payers", and doesn't publish the weights (${a(src(76), 'Bazaar')}). Google, ProRata and Perplexity pick sources inside their own products.`,
+ `from your FORA file and your Exchange, or through a Broker that orders offers by relevance (see ${a('#broker-order', 'How a Broker orders offers')}). No index spans every FORA site.`,
+ ) },
+];
+const QUESTIONS_REST = [
+ { title: 'Who is the buyer, provably?', text: qa(
+ `most use an account and an API key; TollBit and MonetizationOS also match user agents. Cloudflare, AWS and Akamai check Web Bot Auth signatures, and Parallel's crawlers sign with it ${cite(44, 57, 58, 86)}.`,
+ `a signature on every request, checked against a key on the buyer's own domain. Cloudflare's bot verification doesn't recognize a FORA agent: it "will fail to verify a message if" it uses the later drafts' header form, which FORA's SDKs send (${a(src(57), 'Cloudflare')}). FORA's edge function checks the signed link itself, and your Cloudflare rules must let signed requests through.`,
+ ) },
+ { title: 'What lets a given use happen?', text: qa(
+ 'a pass only TollBit can check; a signed price header Cloudflare checks; an x402 payment AWS verifies; an API key.',
+ 'a signed link that expires and can be bound to the agent\'s key. Any edge holding the Exchange\'s public key can check it.',
+ ) },
+ { title: 'Who sets the price, in what unit?', text: qa(
+ 'you, at bot paywalls, Dappier, Cashmere, Snowflake and CCC\'s per-order sales. The channel itself, at answer engines, Google, Valyu, AskNews and Factiva. The buyer, in Pay Per Use. Units run from a page to a query, a summary, a share or a year.',
+ `you, per term. A term is free, flat or per unit, limited by use, country and buyer type, with caps (${a(src(5), 'Licensing terms')}). FORA has no revenue-share price. A buyer that already pays you a share under a contract buys under a deal term (${a(src(5, '#revenue-share-arrangements'), 'Revenue share')}).`,
+ ) },
+ { title: 'How does money move?', text: qa(
+ 'TollBit, Cloudflare, CCC and Snowflake collect from the buyer and pay you; Google pays you from its own programs; the rails move money per request.',
+ FORA_CELLS.money,
+ ) },
+ { title: 'What do you see?', text: qa(
+ `TollBit and Cloudflare show bot traffic ${cite(31, 37)}; AWS lists each settlement; Google shows one monthly earnings number; ProRata sends monthly statements per article.`,
+ 'each sale, in the publisher console your Exchange operator runs. No protocol call returns your sales.',
+ ) },
+ { title: 'What if the numbers are wrong?', text: qa(
+ 'TollBit and Cloudflare bill only on a delivered page and document no dispute call; ProRata grants an audit right.',
+ FORA_CELLS.disputes,
+ ) },
+];
+
+const SETUP = [
+ { title: 'Register your domain.', text: `On the publisher console, name your domain and the paths you license, then choose your terms (${a(src(1), 'Publisher onboarding')}).` },
+ { title: 'Serve your FORA file.', text: 'Serve it unchanged on your site. A later change to it needs no redeploy.' },
+ { title: 'Add the edge function.', text: 'The console builds it for Cloudflare or CloudFront, and it sits beside what your CDN already runs.' },
+ { title: 'Let two kinds of request through your firewall.', text: `Signed requests, and requests for your FORA file, must get past your WAF and bot product (${a(src(3, '#keeping-cdn-bot-protection-on'), 'Keeping CDN bot protection on')}).` },
+ { title: 'Keep a sitemap.', text: 'The operator imports your pages from it.' },
+];
+
+const BUY_STEPS = [
+ { text: 'You set your terms with your Exchange, and your FORA file names the Exchanges that sell your pages.' },
+ { text: `An AI agent signs every request with one key, published on its own domain. An assistant on MCP gets one from FORA's ${a(src(6), 'Identity Service')}.` },
+ { text: 'The agent asks a Broker, with a list of URLs or in plain words. For each URL, the Broker reads your FORA file and asks one of your Exchanges that it trusts.' },
+ { text: 'Your Exchange answers with offers it signed, each carrying your terms and price.' },
+ { text: 'The Broker checks every signature and returns the offers, grouped by URL. The agent chooses.' },
+ { text: `The agent signs one acceptance over every offer it picks, across Exchanges. The Broker splits the purchase per Exchange (${a(src(7, '#buying-through-the-broker'), 'Broker')}).` },
+ { text: 'Each Exchange charges the agent\'s own account and issues a signed link. The agent fetches the page from your CDN, which checks the link. The Broker never charges anyone and never sees the content.' },
+ { text: `Your Exchange operator pays you under your agreement. ${SEE_ABOUT}` },
+];
+
+const SELL_THROUGH = [
+ { title: 'Marketplaces with their own catalog.', text: `A retrieval marketplace, a collective or a rights agency can run a FORA Exchange and Broker over its own catalog and keep its search (${a(src(8), 'For content marketplaces')}).` },
+ { title: 'Buyers that pay a share or a fee.', text: `An AI company that pays you a share or a fee under a deal could buy as a FORA agent, under a deal term (${a(src(5, '#revenue-share-arrangements'), 'Revenue share')}).` },
+ { title: 'Rights agencies and rails.', text: 'A CCC order maps almost field for field to a FORA term; CCC documents no buyer API. An Exchange operator can collect over x402 or fund an agent\'s account through Skyfire; settlement stays outside the protocol.' },
+];
+
+const TODO = [
+ { title: 'Ask who has paid.', text: 'Ask each channel for named buyers and the totals they paid.' },
+ { title: 'Ask for the fee.', text: `TollBit's, Cloudflare's and Microsoft's aren't published. Ask what a buyer pays in total and what you keep. FORA itself is free, open-source software; for the fee at FORA\'s publisher console, see ${ABOUT}.` },
+ { title: 'List what you hand over.', text: 'A subdomain, your logs with your readers\' requests in them, a copy of your archive, a CDN rule. Ask how many engineer-days each one takes.' },
+ { title: 'Keep one price per page.', text: 'If a page is on sale in two places, know both prices.' },
+ { title: 'Order your CDN rules.', text: 'Let each partner\'s fetcher and the search crawlers through, and keep each paywall\'s rule off the others\' paying buyers.' },
+ { title: 'Sell to agents as well.', text: 'Register with an Exchange operator, serve your FORA file and install the edge function on Cloudflare, Fastly or CloudFront. Agents can then buy under your terms beside every channel above.' },
+];
+
+const LEGAL = [
+ { title: 'Exclusivity.', text: `Check whether a grant includes training, and whether it is exclusive. An exclusive training grant stops you selling training anywhere else. ProRata's MSA is non-exclusive (${a(src(80), 'ProRata MSA')}).` },
+ { title: 'Room for another channel.', text: 'Microsoft\'s PCM runs beside FORA on one site only if your PCM contract allows it.' },
+ { title: 'A record of each FORA sale.', text: `With each purchase the buyer signs an acceptance of your terms, and the publisher console can check that signature (${a(src(1, '#after-setup'), 'Publisher onboarding')}).` },
+];
+
+// Deep pages: the series' pages (their index summaries), with the collective
+// page from the Resources list, in the copy's order.
+const collective = RESOURCES.find((r) => r.href === '/resources/transparent-collective/');
+const series = SERIES.filter((p) => p.href !== HREF);
+const DEEP = [
+ ...series.filter((p) => p.href.startsWith(HREF)),
+ ...(collective ? [collective] : []),
+ ...series.filter((p) => !p.href.startsWith(HREF)),
+].map(({ href, title, summary }) => ({ href, title, summary }));
+
+const updated = resourceUpdated(HREF) ?? SERIES.find((p) => p.href === HREF)?.updated;
+---
+
+
+ See the map
+ What FORA takes
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
The map's "Who pays today" column has the rest, in each channel's own words.
+
+
+
+
+
Reading note. The rows compare channels built for different jobs: a rail carries no license, and a collective doesn't set your price. "Who pays today", "Who sets the price" and "What your engineers change" compare like for like. A dash means the channel has nothing to state there. Figures from a vendor are the vendor's own and directional.
+
+
Court settlements price no future use, so they fit none of the three.
+
+
Announced, not running. Next Net and Sundial's SAIL ledger (2026-07-14) names no buyer and publishes no specification ${cite(99)}. AWS hasn't announced the content marketplace The Information reported in February 2026 ${cite(100)}.`} />
+
+ What your engineers change, in full
+
+ {ENGINEERS_FULL.map((it) =>
{it.title}
)}
+
+
+
+
+
+
+
+
Bot paywalls: the bot pays at your door
+
tollbit.news.example.com, which TollBit runs. By default a bot without TollBit's pass gets HTTP 402, "payment required". The buyer pays your rate plus "a small transaction fee on top of the rates you set", whose size isn't published (${a(src(27), 'Setting rates')}).`} />
+
+
+
+
+
+
+
+ {KINDS.map((k) => (
+
+
{k.title}
+ {k.paras.map((p) => )}
+
+ ))}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Beside TollBit. TollBit says another marketplace can run "in parallel" (${a(src(32), 'TollBit')}). Let TollBit's fetcher through FORA's edge, and keep TollBit's rule off requests that carry a FORA signed link. You write that order; FORA ships no combined rule.`} />
+
Beside Cloudflare. FORA's edge runs as a Worker after Cloudflare's own rules.`} />
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
How a Broker orders offers
+
+
+
+
Your TollBit rates, shown to FORA agents
+
+
+
How other channels can use FORA
+
+
+
+
What FORA can't do with them
+
Settling a FORA sale through TollBit or Pay Per Crawl. TollBit's API bills at TollBit's rates, and Pay Per Crawl charges when a crawler fetches a page. Neither one's docs describe a way to bill an amount for a sale made elsewhere.`} />
+
Answer engines' and Google's content as offers. ProRata, Perplexity, Microsoft and Google license your content for their own products, and Microsoft for the partners it brings in. None offers it to any other agent.`} />
+
+
+
+
+
+
+
+
+ What is FORA?
+ Publisher onboarding
+ For content marketplaces
+
+
+
+
diff --git a/website/src/pages/resources/content-marketplaces/microsoft-pcm.astro b/website/src/pages/resources/content-marketplaces/microsoft-pcm.astro
new file mode 100644
index 00000000..b8012a19
--- /dev/null
+++ b/website/src/pages/resources/content-marketplaces/microsoft-pcm.astro
@@ -0,0 +1,436 @@
+---
+// Resource: Microsoft Publisher Content Marketplace (PCM) and FORA, one of the
+// deep pages of the content-marketplaces series (hub:
+// /resources/content-marketplaces/). Laid out in the vendor page template, as
+// the TollBit page: hero, at a glance, the short answer, how you're paid, key
+// facts, who has signed and been paid, what it takes on your site, search
+// traffic, where PCM is ahead, the eight questions, running both, what to do,
+// and About FORA with the sources and the series navigation.
+//
+// Shared texts (About FORA, the definitions, the search callout, the table
+// questions, the GlanceCard rows and FORA's fixed cells) come from
+// src/data/content-marketplaces.ts. A claim cites its source inline, as a
+// link, or with cite(N), which links to entry N of the numbered SourceList at
+// the foot; SOURCES below is numbered as the copy numbers it, so an entry's
+// position is its number. Text with links is HTML and goes through set:html.
+import LandingLayout from '../../../layouts/LandingLayout.astro';
+import AudienceHero from '../../../components/landing/AudienceHero.astro';
+import Band from '../../../components/landing/Band.astro';
+import GlanceCard from '../../../components/landing/GlanceCard.astro';
+import CardGrid from '../../../components/landing/CardGrid.astro';
+import Callout from '../../../components/landing/Callout.astro';
+import StepList from '../../../components/landing/StepList.astro';
+import Figure from '../../../components/landing/Figure.astro';
+import KeyFacts from '../../../components/landing/KeyFacts.astro';
+import PullQuote from '../../../components/landing/PullQuote.astro';
+import Split from '../../../components/landing/Split.astro';
+import PointList from '../../../components/landing/PointList.astro';
+import CompareTable from '../../../components/landing/CompareTable.astro';
+import CodeCard from '../../../components/landing/CodeCard.astro';
+import SourceList from '../../../components/landing/SourceList.astro';
+import SeriesNav from '../../../components/landing/SeriesNav.astro';
+import ClosingCta from '../../../components/landing/ClosingCta.astro';
+import { resourceUpdated } from '../../../data/resources';
+import {
+ ABOUT_FORA, DEFINITIONS, SEARCH_FORA, FORA_CELLS, SEE_ABOUT, compareRows, glanceGroups,
+} from '../../../data/content-marketplaces';
+import { cite, statusTag } from '../../../lib/resource-markup';
+
+const HREF = '/resources/content-marketplaces/microsoft-pcm/';
+const IMG = '/resources/content-marketplaces/microsoft-pcm';
+
+// Every external source the page links.
+const M = {
+ launch: 'https://about.ads.microsoft.com/en/blog/post/february-2026/building-toward-a-sustainable-content-economy-for-the-agentic-web',
+ product: 'https://preview-about.ads.microsoft.com/en/solutions/technology/publisher-content-marketplace',
+ bing: 'https://blogs.bing.com/search/2026/10/your-content-your-choice-clearer-preferences-for-training-and-grounding/',
+ kolar: 'https://digiday.com/media/qa-nikhil-kolar-vp-microsoft-ai-scales-its-click-to-sign-ai-content-marketplace/',
+ yahoo: 'https://www.yahooinc.com/press/introducing-yahoo-scout-a-new-ai-answer-engine',
+ sej: 'https://www.searchenginejournal.com/googles-ai-payment-pilot-vs-cloudflare-and-microsoft-models/589636/',
+ gannett: 'https://www.editorandpublisher.com/stories/gannett-announces-third-quarter-2025-results-new-ai-licensing-deal-with-microsoft-and-updated,258540',
+ iac: 'https://ir.iac.com/static-files/e2666073-0141-41f8-a45d-77ea800a65be',
+ ap: 'https://digiday.com/media/media-briefing-associated-press-deal-cements-microsofts-quiet-rise-in-ai-licensing',
+ bi: 'https://talkingbiznews.com/media-news/business-insider-joining-microsoft-content-marketplace/',
+ verge: 'https://www.theverge.com/news/873296/microsoft-publisher-content-marketplace-ai-licensing',
+};
+// FORA's docs, as the copy links them.
+const D = {
+ onboarding: '/getting-started/publisher-onboarding/',
+ afterSetup: '/getting-started/publisher-onboarding/#after-setup',
+ serveFile: '/getting-started/publisher-onboarding/#serve-your-discovery-file',
+ deployment: '/components/edge-function/deployment/',
+ providersDocs: '/components/edge-function/deployment/#the-providers-own-documents',
+ botDetection: '/components/edge-function/bot-detection/',
+ keepBotProtection: '/components/edge-function/bot-detection/#keeping-cdn-bot-protection-on',
+ whatIsFora: '/getting-started/what-is-fora/',
+ howItWorks: '/getting-started/what-is-fora/#how-it-works',
+ edgeOverview: '/components/edge-function/overview/',
+ terms: '/protocol/licensing-terms/',
+ metering: '/protocol/licensing-terms/#metering-modes',
+ revenueShare: '/protocol/licensing-terms/#revenue-share-arrangements',
+ broker: '/components/broker/overview/',
+ auth: '/protocol/authentication/',
+};
+const a = (href: string, text: string) => `${text}`;
+const NP = statusTag('not-public');
+
+const GLANCE = glanceGroups({
+ payer: [`Microsoft, for Copilot's use: "paid on delivered value" ${cite(1)}. When, and who pays for other buyers' use: ${NP}`, FORA_CELLS.payer],
+ fees: [`For Copilot's use, Microsoft is the buyer ${cite(1)}.`, FORA_CELLS.fees],
+ reporting: [
+ `"Usage-based reporting" ${cite(1)}. What a report contains: ${NP}`,
+ `Each sale, with gross, the Exchange fee and net, in your operator's publisher console (${a(D.afterSetup, 'Publisher onboarding')}).`,
+ ],
+ contract: [
+ `With Microsoft, negotiated one by one in the pilot. The standard contract wasn't final in February 2026 ${cite(12)}.`,
+ `Your agreement with your Exchange operator (${a(D.deployment, 'Deployment')}). Your terms go into each offer the Exchange signs.`,
+ ],
+ status: [`${statusTag('pilot')} No payment is public ${cite(1)}.`, `${statusTag('live')} ${SEE_ABOUT}`],
+ dns: [`No change documented ${cite(1, 2)}.`, `No record to add (${a(D.deployment, 'Deployment')}).`],
+ cdn: [
+ `No change documented ${cite(1, 2)}.`,
+ 'A drop-in edge function. The rest depends on your own setup; FORA adds to what you run rather than replacing it.',
+ ],
+ firewall: [
+ `No bot rule documented. What your CDN blocks stays your decision ${cite(1, 2)}.`,
+ `Your firewall and bot product run first. They must let signed requests and your FORA file through (${a(D.keepBotProtection, 'Bot detection')}).`,
+ ],
+ site: [
+ `You hand over content through "the pipes" Microsoft built ${cite(12)}. Feed, API or crawl: ${NP}`,
+ `One file on your site, fora.json, naming your Exchanges. The operator imports your pages from your sitemap (${a(D.afterSetup, 'Publisher onboarding')}).`,
+ ],
+ search: [
+ `Bing's crawl is separate from PCM ${cite(3)}. How Bing's settings affect a PCM license: ${NP} See Search traffic, below.`,
+ `Googlebot, Bingbot and Applebot always pass ${cite(6)}. See Search traffic, below.`,
+ ],
+ upkeep: [
+ `How often you send content, and who keeps it current: ${NP}`,
+ `A change to your FORA file takes effect with nothing to redeploy (${a(D.providersDocs, 'Deployment')}).`,
+ ],
+ leaving: [`"Participation is voluntary" ${cite(1)}. How you leave: ${NP}`, FORA_CELLS.leaving],
+});
+
+const SHORT_ANSWER = [
+ { title: 'PCM brings a buyer.', text: `Microsoft runs the marketplace and buys through Copilot, and says it is "beginning to onboard" Yahoo ${cite(1)}. You sign a contract and hand over your content. Microsoft documents no change to your site.` },
+ { title: 'FORA brings an open interface.', text: 'Any agent finds your offers, signs for the terms it picks and gets an expiring link for that purchase. You serve one file and install the edge function on your CDN.' },
+ { title: 'They run side by side.', text: 'PCM publishes no API for outside software, so FORA can\'t connect to it without Microsoft. Both can run on one site, if your PCM contract allows it.' },
+];
+
+const PAID_STEPS = [
+ { title: 'You register interest.', text: `Microsoft's product page says: "Register your interest in the Publisher Content Marketplace to receive updates as our work progresses" (${a(M.product, 'product page')}). How publishers are selected: ${NP}` },
+ { title: 'You sign a contract.', text: `"With the pilot, we did individual negotiations with the publishers," Microsoft's Nikhil Kolar told Digiday in February 2026. A standard contract is planned, "but we’ve not finalized it yet" (${a(M.kolar, 'Digiday interview')}).` },
+ { title: 'You hand over your content.', text: `Kolar describes the first phase as getting "the pipes working — how does the content come in". Publishers can "clean up the content", for example an image licensed from someone else ${cite(12)}. By feed, API or crawl: ${NP}` },
+ { title: 'You set the terms.', text: `"Publishers define licensing and usage terms", with no published vocabulary ${cite(1)}. The primary use, in Microsoft's words, is "Grounding and Display": content grounds the answer to a single query, "and that same content is also displayed as a citation URL" in it. "PCM is for Grounding and Display use cases, not training" ${cite(2)}.` },
+ { title: 'Copilot uses it.', text: `Microsoft "started with a focused set of scenarios in enterprise and consumer versions of Microsoft Copilot". It is "beginning to onboard demand partners, including Yahoo" ${cite(1)}. Yahoo said on January 27, 2026, that it "is also joining Microsoft’s Publisher Content Marketplace pilot" (${a(M.yahoo, 'Yahoo')}). Whether Yahoo buys through PCM yet: ${NP}` },
+ { title: 'You\'re paid, and you get a report.', text: `PCM "provides usage-based reporting" ${cite(1)}. What a report contains: ${NP} Search Engine Journal noted the gap: the post "doesn’t say who sets the price or what the reports show" (${a(M.sej, 'SEJ')}).` },
+];
+
+const KEY_FACTS = [
+ { value: '2', text: `versions of Copilot PCM started with, enterprise and consumer ${cite(1)}` },
+ { value: '7', text: `US publishers Microsoft names as co-designers, "among others" ${cite(1)}` },
+ { value: '4', text: `publishers that confirmed their deals in late 2025 ${cite(15, 16, 17, 18)}` },
+ { value: '1', text: `buyer beyond Copilot that Microsoft names: Yahoo, which it is onboarding ${cite(1)}` },
+];
+
+const SITE_PCM = [
+ { title: 'How content gets in:', text: `through "the pipes" Microsoft built ${cite(12)}. Feed format, ingestion API, CMS plugin or crawler: ${NP}` },
+ { title: 'DNS and CDN:', text: `no change documented. The product page promises "a single, trusted path to make content available for AI use cases while preventing unwanted crawling" ${cite(2)}. It doesn't say how.` },
+ { title: 'Who gets in:', text: `Microsoft onboards the other buyers, such as Yahoo ${cite(1)}. How your contract limits them: ${NP} PCM documents no bot rule, so what your CDN blocks stays your decision.` },
+ { title: 'What a bot gets back:', text: 'PCM documents no answer to a bot, neither a redirect nor a payment-required response.' },
+ { title: 'Logs:', text: `none documented. You get "usage-based reporting" ${cite(1)}.` },
+];
+const SITE_FORA = [
+ { title: 'How content gets in:', text: `it stays on your site. The Exchange reads your sitemap, and you set terms in its operator's console. Or you push the catalog to it (${a(D.howItWorks, 'What is FORA?')}).` },
+ { title: 'DNS and CDN:', text: `you serve one file that names your Exchanges, and install the edge function on your CDN (${a(D.deployment, 'Deployment')}).` },
+ { title: 'Who gets in:', text: `you choose the paths you license. There, the edge function refuses AI bots that come without a signed link. Readers and search crawlers pass as before ${cite(6)}.` },
+ { title: 'What a bot gets back:', text: `a refusal that tells the bot where to buy ${cite(8)}.` },
+ { title: 'Logs:', text: `each refusal states its reason, the same reason the edge function logs ${cite(8)}. The Exchange records each sale.` },
+];
+
+const ENGINEER_STEPS = [
+ { title: 'Serve the file.', text: `Your operator's console gives you the file that names your Exchanges. Serve it unchanged (${a(D.serveFile, 'Publisher onboarding')}).` },
+ { title: 'Add the edge function.', text: `The console builds it for your domain, for Cloudflare or CloudFront, and it sits beside what your CDN already runs (${a(D.deployment, 'Deployment')}).` },
+ { title: 'Let signed requests through.', text: `On most CDNs, the firewall and bot product run before the edge function, so a bot rule there can refuse an agent that has already bought. Let signed requests through it (${a(D.keepBotProtection, 'Bot detection')}).` },
+ { title: 'How the edge function decides.', text: `On the paths you license, it matches the name each bot sends, its User-Agent, against FORA's published list of AI bots. It never matches Googlebot, Bingbot or Applebot ${cite(6)}.` },
+ { title: 'What a refused bot gets.', text: `A 403, with a header that points it to the file naming your Exchanges ${cite(8)}. There is no 402: the agent buys at the Exchange, before it fetches.` },
+];
+
+const CAUTIONS = [
+ { title: 'Blocking Bingbot takes you out of Bing search.', text: 'Bing\'s own grounding refusal doesn\'t.' },
+ { title: 'Grounding is refused page by page.', text: 'You set it with a meta tag or a response header: "There is no site-wide robots.txt grounding preference." Training can be refused per site or path in robots.txt, with the shared AI-training group that other crawlers can honor too, or per page.' },
+ { title: 'Bing has to crawl to see your choice.', text: '"Microsoft must crawl the page to discover the setting."' },
+ { title: 'NoArchive and NoCache stop counting.', text: 'After March 1, 2027, Bing no longer reads these older tags for training and grounding. If you use them to stay out of AI answers, also set a temporary preference in Bing Webmaster Tools before that date.' },
+];
+
+const AHEAD = [
+ { title: 'The buyer comes with the marketplace.', text: `PCM's operator is also its first buyer. Copilot comes in two versions, "the M365 copilot from the enterprise side, and the MAI copilot from the consumer side" ${cite(12)}. On FORA, any agent can buy.` },
+ { title: 'No change to your site that Microsoft documents.', text: `PCM needs a contract and your content: "Instead of blocking bots or negotiating one-off deals, tap into a transparent marketplace" ${cite(2)}. FORA asks for a file and an edge function, as above.` },
+ { title: 'A price set by what the answer needed.', text: `Kolar: "Sometimes, the value may be: ‘I just need to know the score.’" For a source that is "not substitutable", a buyer is "willing to pay a lot more" ${cite(12)}. PCM is designed to price a use by its worth in the answer. FORA can't: its price is fixed at the sale, and a usage report "never re-prices a purchase" (${a(D.metering, 'Licensing terms')}).` },
+ { title: 'Reports from inside the AI product.', text: `Microsoft runs Copilot, so its reports can come from the answer itself. Kolar: "we wanted to build a feedback loop from the demand side on how the content was used" ${cite(12)}. FORA can't report that. It records each sale, and of the use in an answer it knows only what the agent reports (${a(D.howItWorks, 'What is FORA?')}, step 5).` },
+ { title: 'A use that includes the citation.', text: `Grounding and Display ties one answer to one shown link. FORA can state the same duty in a term, as below, but nothing in FORA observes that the link was shown. For an outside buyer such as Yahoo, how PCM learns of the citation: ${NP}` },
+ { title: 'Deals you\'ve already made.', text: `Kolar calls it "bring-your-own-license": PCM "can become the infrastructure to deliver any license that you have negotiated outside the marketplace". "Some of the larger partners we work with already have a licensing deal," he says ${cite(12)}. PCM is built to carry such deals, to buyers Microsoft contracts. FORA carries them as a deal term (${a(D.revenueShare, 'Licensing terms')}).` },
+];
+
+const COMPARE = compareRows([
+ [
+ `In your PCM contract: "publisher-defined content and license terms". Nothing on your site says so to anyone else ${cite(2)}.`,
+ `Your FORA file names the Exchanges that sell your content, and you set terms per resource in your operator's publisher console ${cite(7)}.`,
+ ],
+ [
+ `"AI builders discover and license content for specific grounding scenarios" ${cite(1)}. How: ${NP}`,
+ `An agent asks your Exchange, or a Broker. Each Broker operator chooses its search, and each agent chooses its Broker ${cite(10)}.`,
+ ],
+ [
+ `Copilot, Microsoft's own product. Microsoft is "beginning to onboard" Yahoo ${cite(1)}. Agent identification: ${NP}`,
+ `Each request carries a signature, checked against the signer's published key (Web Bot Auth, the IETF working group's -00 draft) ${cite(11)}.`,
+ ],
+ [
+ `The contract. Pilot contracts were negotiated one by one; the standard contract was not final in February 2026 ${cite(12)}.`,
+ `A signed offer the agent accepts and signs, and a link that expires, checked at your CDN ${cite(7)}.`,
+ ],
+ [
+ `Paid "on delivered value" ${cite(1)}. The unit and who sets the number: ${NP}`,
+ `You: free, per unit (an access, a token, a page) or a flat fee, signed on each offer. Fixed at the sale, by design ${cite(9)}.`,
+ ],
+ [`For Copilot's use, Microsoft is the buyer ${cite(1)}. Who pays for any other partner's use, when, and any minimum: ${NP}`, FORA_CELLS.money],
+ [
+ `"usage-based reporting" on how content is "valued, cited, and used" ${cite(1, 2)}. What a report contains: ${NP}`,
+ `The operator's console: your sales, with gross, fee and net. It is each operator's own; no protocol call returns your sales ${cite(4)}.`,
+ ],
+ [`No dispute process is described ${cite(1, 2, 12)}.`, FORA_CELLS.disputes],
+]);
+
+const WORKS = [
+ { title: 'Run both.', text: `"Participation is voluntary," Microsoft says ${cite(1)}. For exclusivity, see For legal, below. PCM stays a contract, and the file on your site names your FORA Exchanges.` },
+ { title: 'Sell the same use to every other agent.', text: 'Grounding and Display is one FORA term. In words: AI answers are allowed and training isn\'t, the article is shown as a cited link, and each access costs $0.05.' },
+];
+
+// The term as it reads on the wire: field, then value.
+const TERM_LINES: [string, string][] = [
+ ['restrictions:', 'FUNCTION permitted: ai-input prohibited: ai-train'],
+ ['obligations:', 'ATTRIBUTION, on use:\n"Show the article as a cited link to news.example.com"'],
+ ['pricing:', 'PER_UNIT unit: accesses rate: "0.05" currency: USD'],
+];
+
+const TODO = [
+ { title: 'Ask how you join.', text: 'Microsoft\'s product page has an interest form, and doesn\'t say how publishers are selected.' },
+ { title: 'Ask for the unit and the price.', text: 'Microsoft hasn\'t published what it pays per use, or who sets the number. Ask both, in writing.' },
+ { title: 'Ask how your content gets in.', text: 'Microsoft hasn\'t published the route: a feed, an API or a crawl. Ask what you send, how often, and who keeps it up to date.' },
+ { title: 'Ask what the report shows.', text: `Microsoft promises insight into how content is "valued, cited, and used" ${cite(2)}. Ask for a sample report.` },
+ { title: 'Ask how Bing\'s settings affect your license.', text: 'If you refuse grounding in Bing, do you lose PCM money? If you\'re in PCM, can you still refuse it? Ask before you choose a setting.' },
+ { title: 'Sell to other agents too.', text: `Name your FORA Exchanges in fora.json and put a grounding term with a citation duty on your articles. FORA itself is free, open-source software. For the publisher console's fees, ${SEE_ABOUT.replace('See', 'see')}` },
+];
+
+const LEGAL = [
+ { title: 'Who you sign with.', text: `Microsoft. Pilot contracts were "individual negotiations"; a standard contract is planned, "but we’ve not finalized it yet" ${cite(12)}.` },
+ { title: 'What you license.', text: `"PCM is for Grounding and Display use cases, not training" ${cite(2)}. The terms are in your contract; Microsoft publishes no vocabulary for them.` },
+ { title: 'Rights you may not hold.', text: `Agency photos and syndicated stories may not be yours to hand over. Kolar gives a licensed-in image as the example ${cite(12)}. List what you can't license before you sign.` },
+ { title: 'Exclusivity.', text: `"Participation is voluntary" ${cite(1)}. Whether the contract lets you sell the same articles through FORA or anyone else: ${NP} Ask before you do.` },
+ { title: 'Bing\'s settings aren\'t a license.', text: `Bing says its preferences "do not establish licensing or payment arrangements" ${cite(3)}. How they interact with a PCM license: ${NP}` },
+ { title: 'Law, venue, disputes and exit.', text: `${NP} No dispute process is described.` },
+ { title: 'Changes.', text: `PCM is a pilot. Kolar: "what we launch may not be what we end up with" ${cite(12)}.` },
+];
+
+// The SourceList, numbered as the copy numbers it: an entry's position,
+// counted across the groups, is the N in cite(N).
+const fora = (title: string, href: string) => ({ title, publisher: 'FORA documentation', href, checked: '2026-11-15' });
+const SOURCES = [
+ {
+ name: 'Microsoft',
+ items: [
+ { title: 'Building toward a sustainable content economy for the agentic web', publisher: 'Microsoft Advertising', href: M.launch, checked: '2026-11-14' },
+ { title: 'Publisher Content Marketplace', publisher: 'Microsoft Advertising, served on Microsoft\'s preview host (preview-about.ads.microsoft.com)', href: M.product, checked: '2026-11-15' },
+ { title: 'Your content, your choice: clearer preferences for training and grounding', publisher: 'Bing', href: M.bing, checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'FORA',
+ items: [
+ fora('Publisher onboarding', D.onboarding),
+ fora('Deployment', D.deployment),
+ fora('Bot detection', D.botDetection),
+ fora('What is FORA?', D.whatIsFora),
+ fora('Edge function overview', D.edgeOverview),
+ fora('Licensing terms', D.terms),
+ fora('Broker', D.broker),
+ fora('Authentication', D.auth),
+ ],
+ },
+ {
+ name: 'Press, filings and others',
+ items: [
+ { title: 'Q&A: Nikhil Kolar, vp Microsoft AI', publisher: 'Digiday', href: M.kolar, checked: '2026-11-14' },
+ { title: 'Introducing Yahoo Scout', publisher: 'Yahoo', href: M.yahoo, checked: '2026-11-14' },
+ { title: 'Google\'s AI payment pilot vs. Cloudflare and Microsoft models', publisher: 'Search Engine Journal', href: M.sej, checked: '2026-11-14' },
+ { title: 'Gannett announces third quarter 2025 results', publisher: 'Editor and Publisher', href: M.gannett, checked: '2026-11-14' },
+ { title: 'Third quarter 2025 results', publisher: 'IAC', href: M.iac, checked: '2026-11-14' },
+ { title: 'Associated Press deal cements Microsoft\'s quiet rise in AI licensing', publisher: 'Digiday', href: M.ap, checked: '2026-11-14' },
+ { title: 'Business Insider joining Microsoft content marketplace', publisher: 'Talking Biz News', href: M.bi, checked: '2026-11-14' },
+ { title: 'Microsoft\'s Publisher Content Marketplace', publisher: 'The Verge', href: M.verge, checked: '2026-11-14' },
+ ],
+ },
+];
+---
+
+
+ Where PCM is ahead
+ Use both
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Microsoft publishes no count of publishers. Neither Microsoft nor a publisher has said a payment was made.
+
+
+
+
+
+
PCM
+
+
+
+
FORA
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
PCM documents nothing that touches search crawlers. Bing's crawl, above, is a separate route into Copilot.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Works today
+
+
+
+ Grounding and Display as one FORA term.
+
+
+
Doesn't fit
+
Paying FORA sales through PCM. PCM is a marketplace with Microsoft as the payer, and no payment service for others. On FORA, the Exchange operator bills the buyer and pays you (${a(D.howItWorks, 'What is FORA?')}).`} />
+
Declaring PCM in a file. Microsoft publishes no file or tag that says a site is in PCM. Asked whether RSL could work with PCM, Microsoft's spokesperson didn't have "anything specific to share today" (${a(M.verge, 'The Verge')}, February 3, 2026). Say it on your licensing page, in words.`} />
+
+
+
+
+
+
+
+
+
+
+
Sources
+
Sources are checked on the date of publication.
+
+
+
+
+
+
+ Licensing terms
+ Run your own marketplace
+ All content marketplaces
+ What FORA takes on your CDN
+
+
+
+
diff --git a/website/src/pages/resources/content-marketplaces/perplexity.astro b/website/src/pages/resources/content-marketplaces/perplexity.astro
new file mode 100644
index 00000000..77701219
--- /dev/null
+++ b/website/src/pages/resources/content-marketplaces/perplexity.astro
@@ -0,0 +1,389 @@
+---
+// Resource: Perplexity's Publishers' Program and Comet Plus, and FORA, one of the deep pages of the
+// content-marketplaces series (hub: /resources/content-marketplaces/). Laid
+// out in the vendor page template, as the TollBit page is: hero, at a glance,
+// the short answer, how Comet Plus is meant to pay you, key facts, who has
+// been paid, what it takes on your site, what can go wrong, where Perplexity
+// is ahead, the eight questions, running both, what to do, the background,
+// the sources, and About FORA with the series navigation.
+//
+// Shared texts (About FORA, the definitions, the search callout, the table
+// questions, the GlanceCard rows and FORA's fixed cells) come from
+// src/data/content-marketplaces.ts. A claim cites its source inline, as a
+// link, or with cite(N), which links to entry N of the numbered SourceList;
+// SOURCES below is numbered in order of first citation, so an entry's
+// position is its number. Text with links is HTML and goes through set:html.
+import LandingLayout from '../../../layouts/LandingLayout.astro';
+import AudienceHero from '../../../components/landing/AudienceHero.astro';
+import Band from '../../../components/landing/Band.astro';
+import GlanceCard from '../../../components/landing/GlanceCard.astro';
+import CardGrid from '../../../components/landing/CardGrid.astro';
+import Callout from '../../../components/landing/Callout.astro';
+import StepList from '../../../components/landing/StepList.astro';
+import Figure from '../../../components/landing/Figure.astro';
+import KeyFacts from '../../../components/landing/KeyFacts.astro';
+import PullQuote from '../../../components/landing/PullQuote.astro';
+import Split from '../../../components/landing/Split.astro';
+import PointList from '../../../components/landing/PointList.astro';
+import CompareTable from '../../../components/landing/CompareTable.astro';
+import StatusTag from '../../../components/landing/StatusTag.astro';
+import SourceList from '../../../components/landing/SourceList.astro';
+import SeriesNav from '../../../components/landing/SeriesNav.astro';
+import ClosingCta from '../../../components/landing/ClosingCta.astro';
+import { resourceUpdated } from '../../../data/resources';
+import {
+ ABOUT_FORA, DEFINITIONS, SEARCH_FORA, FORA_CELLS, SEE_ABOUT, compareRows, glanceGroups,
+} from '../../../data/content-marketplaces';
+import { cite, statusTag } from '../../../lib/resource-markup';
+
+const HREF = '/resources/content-marketplaces/perplexity/';
+const IMG = '/resources/content-marketplaces/perplexity';
+
+// Every external source the page links.
+const X = {
+ comet: 'https://www.perplexity.ai/hub/blog/introducing-comet-plus',
+ partners: 'https://www.perplexity.ai/hub/blog/announcing-comet-plus-launch-partners',
+ program: 'https://www.perplexity.ai/hub/blog/introducing-the-perplexity-publishers-program',
+ pricing: 'https://www.perplexity.ai/hub/pricing',
+ crawlers: 'https://docs.perplexity.ai/docs/resources/perplexity-crawlers',
+ robots: 'https://www.perplexity.ai/help-center/en/articles/10354969-how-does-perplexity-follow-robots-txt',
+ bullying: 'https://www.perplexity.ai/hub/blog/bullying-is-not-innovation',
+ keys: 'https://www.perplexity.ai/.well-known/http-message-signatures-directory',
+ dowJones: 'https://www.perplexity.ai/hub/blog/about-the-dow-jones-lawsuit',
+ searchApi: 'https://www.perplexity.ai/hub/legal/perplexity-api-terms-of-service-search',
+ plans: 'https://www.perplexity.ai/help-center/en/articles/11187416-which-perplexity-subscription-plan-is-right-for-you',
+ digiday: 'https://digiday.com/media/how-perplexity-new-revenue-model-works-according-to-its-head-of-publisher-partnerships/',
+ gigazine: 'https://gigazine.net/gsc_news/en/20260219-perplexity-drops-ads-hurt-trust-ai',
+ hurdles: 'https://digiday.com/media/media-briefing-here-are-the-hurdles-to-perplexitys-pitch-as-the-publisher-friendly-llm',
+ cnn: 'https://storage.courtlistener.com/recap/gov.uscourts.nysd.664916/gov.uscourts.nysd.664916.1.0_1.pdf',
+ nyt: 'https://www.sec.gov/Archives/edgar/data/71691/000007169126000041/nyt-20260930.htm',
+ pressGazette: 'https://pressgazette.co.uk/news/perplexity-launches-ai-subscription-revenue-share-scheme-for-publishers/',
+ archive: 'https://web.archive.org/web/20251204132150/https://www.perplexity.ai/hub/blog/announcing-comet-plus-launch-partners',
+ thewrap: 'https://www.thewrap.com/industry-news/public-policy-legal/cnn-sues-perplexity-ai-stealing-news-chatbot-copyright/',
+};
+const a = (href: string, text: string) => `${text}`;
+
+const GLANCE = glanceGroups({
+ payer: [`Perplexity says it pays, from a pool of Comet Plus revenue, by its own counts. Schedule, threshold and currency: Not public ${cite(1)}.`, FORA_CELLS.payer],
+ fees: [`Perplexity keeps "a small portion" for compute; 20% in press reports ${cite(1, 12)}.`, FORA_CELLS.fees],
+ reporting: [
+ `No Comet Plus statement format is published ${cite(1, 2)}. 2024 partners got analytics on how Perplexity cites them ${cite(3)}.`,
+ `Each sale, in your operator's publisher console, with gross, the Exchange fee and net (${a('/getting-started/publisher-onboarding/', 'Publisher onboarding')}).`,
+ ],
+ contract: [
+ `A private contract, after an email to Perplexity. No standard agreement is published ${cite(1)}.`,
+ `Your agreement with your Exchange operator; your terms go into each offer it signs ${cite(21)}.`,
+ ],
+ status: [
+ `${statusTag('not-public')} Comet Plus isn't on Perplexity's ${a(X.pricing, 'pricing page')}. Whether it still pays out isn't public.`,
+ `${statusTag('live')} ${SEE_ABOUT}`,
+ ],
+ dns: [`No change in Perplexity's docs ${cite(5)}.`, `No record to add ${cite(20)}.`],
+ cdn: [
+ `No code in Perplexity's docs ${cite(5)}.`,
+ 'A drop-in edge function. The rest depends on your own setup; FORA adds to what you run rather than replacing it.',
+ ],
+ firewall: [
+ `An allow rule that matches each Perplexity crawler to its own IP ranges ${cite(5)}.`,
+ `Keep your bot protection, and let AI agents that name themselves reach the edge function. Or let signed requests and your discovery file through (${a('/components/edge-function/bot-detection/', 'Bot detection')}).`,
+ ],
+ site: [
+ `No feed, file or API. How your paywall lets a Comet Plus subscriber in: Not public ${cite(5, 6)}.`,
+ `One file that names your Exchanges, and a sitemap the operator imports your pages from ${cite(20)}.`,
+ ],
+ search: [
+ `PerplexityBot builds Perplexity's search index and follows robots.txt ${cite(5, 6)}.`,
+ `Googlebot, Bingbot and Applebot always pass ${cite(22)}. PerplexityBot is refused on licensed paths by default. See Search traffic, below.`,
+ ],
+ upkeep: [
+ `A scheduled job to refresh Perplexity's IP ranges, which Perplexity updates regularly ${cite(5)}.`,
+ `Change your Exchanges by editing your discovery file, with nothing to redeploy ${cite(20)}.`,
+ ],
+ leaving: ['Not public.', FORA_CELLS.leaving],
+});
+
+const SHORT_ANSWER = [
+ { title: 'Perplexity is the buyer.', text: 'Comet Plus is meant to share Perplexity\'s own revenue with the publishers it signs, for use in its answer engine, its Comet browser and Comet Assistant. No other AI company buys publisher content through its programs.' },
+ { status: 'not-public' as const, title: 'Whether Comet Plus still pays isn\'t public.', text: 'Comet Plus isn\'t on Perplexity\'s pricing page or plan guide. CNN, a launch partner, never got from its term sheet to a contract, its complaint says. CNN sued Perplexity on 2026-05-28.' },
+ { title: 'FORA prices the access, on the same site.', text: 'You set the price, the unit and the terms for each page. Any agent can buy, and each purchase is a signed record. You can let Perplexity\'s search crawler in and sell to other agents through FORA on one site.' },
+];
+
+const PAY_STEPS = [
+ { title: 'You email and sign.', text: `"Publishers interested in participating can join the program by emailing us via publishers@perplexity.ai" (${a(X.comet, 'Introducing Comet Plus')}). No standard agreement is published.` },
+ { title: 'Subscribers pay Perplexity.', text: `"Comet Plus will be a $5 standalone subscription. Pro and Max subscribers get Comet Plus included in their memberships" ${cite(1)}.` },
+ { title: 'Subscribers reach your site.', text: `They get "direct access to the sites and content of participating publishers", and their assistants can "complete tasks on those sites" ${cite(1)}. How your paywall recognizes a subscriber isn't public.` },
+ { title: 'Perplexity counts three kinds of traffic.', text: `Comet Plus "allocates revenue to our partners based on three types of internet traffic: human visits, search citations, and agent actions" ${cite(1)}. Digiday describes the visits as "direct visits to publisher's sites by people browsing using Comet". It describes agent actions as content "used to complete tasks by Comet's AI assistant" (${a(X.digiday, 'Digiday')}, published 2025-08-26). Perplexity counts all three in its own products, with no independent count described.` },
+ { title: 'The revenue becomes a pool.', text: `"We're distributing all of that revenue to participating publishers, minus a small portion for Perplexity's compute costs" ${cite(1)}. Digiday reported that "Perplexity keeps 20% of it" and that Perplexity is "opening up a pool of $42.5 million to publishers". Perplexity's own pages give neither figure.` },
+ { title: 'The pool is split by weights you don\'t see.', text: 'Asked whether one kind of use weighs more, Jessica Chan, Perplexity\'s head of publisher partnerships, told Digiday: "The details will be very specific to our partners, but you are thinking about it correctly." The weights aren\'t published.' },
+ { title: 'Perplexity says it pays.', text: 'No schedule, threshold, currency or statement format is published.' },
+];
+
+const KEY_FACTS = [
+ { value: '$5', text: `Comet Plus on its own; Pro and Max subscribers get it included ${cite(1)}` },
+ { value: '3', text: `Kinds of traffic Perplexity counts: human visits, search citations and agent actions ${cite(1)}` },
+ { value: '$42.5 million', text: `The pool Digiday reported; Perplexity's own pages don't give it ${cite(1, 12)}` },
+ { value: 'Not public', text: `The weight on each kind of traffic, per partner ${cite(12)}` },
+];
+
+const ENGINEERS = [
+ { title: 'How your content reaches Perplexity.', text: 'PerplexityBot crawls your pages for Perplexity\'s search index, and "Perplexity partners with third-party crawlers to help build our search index". Perplexity-User fetches a page when a user\'s question needs it. You publish no feed, file or API.' },
+ { title: 'The allow rule.', text: 'If your firewall screens bots, Perplexity asks for an allow rule that matches each crawler\'s user agent to its own published IP ranges. It advises an automated job to keep the ranges current.' },
+ { title: 'On Cloudflare.', text: 'A custom rule set to Skip, placed above the rules that block or challenge, lets both crawlers past them.' },
+ { title: 'With FORA on the same site.', text: `Apply Perplexity's IP-range rules in your CDN's firewall, which runs before FORA's edge function. The edge function knows both crawlers by the names they send, and a name can be copied (${a('/components/edge-function/bot-detection/', 'Bot detection')}).` },
+ { title: 'Nothing is metered at your site.', text: 'Perplexity\'s docs describe no payment response, no redirect and no signed link. It counts visits, citations and agent actions in its own products, and describes no log feed in either direction. Its one word on logs is to "Monitor your logs" to see that its crawlers get through.' },
+];
+
+const CAUTIONS = [
+ { title: 'A blocked page can still show.', text: `If robots.txt blocks a page, Perplexity "may still index the domain, headline, and a brief factual summary" ${cite(6)}.` },
+ { title: 'Changes take time.', text: `A robots.txt change "may take up to 24 hours" to apply ${cite(5)}.` },
+ { title: 'The ranges move.', text: `"These addresses are updated regularly", so an allow rule that isn't refreshed falls behind ${cite(5)}.` },
+ { title: 'Bot Fight Mode stays on.', text: `On Cloudflare, "Bot Fight Mode can't be skipped with WAF custom rules" ${cite(5)}.` },
+];
+
+const AHEAD = [
+ { title: 'You have a buyer from the day you sign.', text: `Perplexity is the buyer. The Financial Times reported "around $200 million in annual revenue" (via ${a(X.gigazine, 'Gigazine')}, published 2026-02-19). A Perplexity executive told it the service has "more than 100 million users", the vendor's own figure. On FORA, any agent your terms allow can buy from your Exchange.` },
+ { title: 'Joining asks only for crawler rules.', text: `Joining is an email and a contract, and the only technical ask Perplexity documents is the crawler rules above. FORA asks you to publish a file that names your Exchanges, sign with an Exchange, set your terms in its console and install its edge function on your CDN (${a('/getting-started/what-is-fora/', 'What is FORA?')}).` },
+ { title: 'It counts traffic FORA can\'t see.', text: `Two of Comet Plus's counts, human visits and agent actions, happen in the reader's browser, as Digiday describes them. Perplexity's position is that "Your AI assistant must be indistinguishable from you" (${a(X.bullying, 'Bullying is not innovation')}, published 2025-11-04). FORA's edge function knows AI bots by the names they send, and a request it doesn't recognize passes as a reader.` },
+ { title: 'Its share counts citations.', text: `Search citations are one of its three counts. FORA fixes the charge when the agent buys, and a usage report "never re-prices a purchase" (${a('/protocol/licensing-terms/#metering-modes', 'Licensing terms')}). A FORA usage report can record a citation, but no FORA pricing unit counts them.` },
+ { title: 'It sends paying readers to your own site.', text: `Comet Plus "is specifically designed to allow publishers to own the direct relationship with their audience" ${cite(2)}. FORA has no reader product. It licenses machine access.` },
+ { title: 'The pool is a set sum, if it still pays.', text: `Publishing executives told Digiday the model "does seem more developed than the ad revenue share program … mostly because the company is setting aside a specific pool of money" (${a(X.hurdles, 'Digiday')}, published 2025-08-28). On FORA, your income is the sum of what agents buy, with no floor.` },
+];
+
+const COMPARE = compareRows([
+ [`In a private contract. Nothing on your site states the terms to anyone else ${cite(1)}.`, `Your fora.json names the Exchanges that sell your content; you set terms per resource in the Exchange operator's console ${cite(21)}.`],
+ [`No outside buyer does. PerplexityBot and third-party crawlers build Perplexity's own index ${cite(5, 6)}.`, `An agent asks your Exchange, or a Broker, which returns your signed offers unchanged (${a('/resources/content-marketplaces/#broker-order', 'how a Broker orders offers')}) ${cite(24)}.`],
+ [`Two crawlers, known by name and published IP ranges ${cite(5)}. Perplexity's crawler docs describe no request signatures, and its site serves no signing-key directory ${cite(5, 8)}. Comet Assistant is meant to be "indistinguishable from you" ${cite(7)}.`, `Each request carries a signature, checked against a key the buyer publishes on its own domain (Web Bot Auth, the IETF working group's -00 draft) ${cite(21, 25)}.`],
+ [`The contract. PerplexityBot follows robots.txt; Perplexity-User, which fetches pages for a user, "generally ignores robots.txt rules" ${cite(5, 6)}.`, `A signed offer the agent accepts and signs, and a signed link that expires, checked at your CDN ${cite(21)}.`],
+ [`Perplexity: the pool, its own cut ("a small portion"; 20% in press reports) and each partner's weights. You set no price ${cite(1, 12)}.`, `You: free, per unit (an access, or text tokens, about ¾ of a word each) or a flat fee, signed on each offer. A revenue share is a term priced at zero that only agents naming the deal see, with a reporting duty (${a('/protocol/licensing-terms/#revenue-share-arrangements', 'Revenue share')}).`],
+ [`Perplexity pays from the pool, by its own account. Schedule, threshold and currency aren't published ${cite(1)}.`, FORA_CELLS.money],
+ [`2024 partners got analytics on how Perplexity cites them ${cite(3)}. No Comet Plus statement format is published ${cite(1, 2)}.`, `Each sale, with its signed offer and the agent's signed acceptance, in your Exchange operator's console (FORA Server's shows gross, fee and net). No protocol call returns your sales ${cite(20)}.`],
+ [`No dispute process is published. Publishers' disagreements with Perplexity have gone to court; see For legal ${cite(9, 15, 16)}.`, FORA_CELLS.disputes],
+]);
+
+const WORKS_TODAY = [
+ { title: 'Run both.', text: 'Your fora.json names your FORA Exchanges. Your Perplexity deal stays a private contract. Check it for exclusivity first; see For legal.' },
+ { title: 'Choose which Perplexity crawler gets in.', text: `On the paths you license, FORA's edge function refuses Perplexity-User without a signed link, as it does every bot that fetches a page because a user asked (${a('/components/edge-function/bot-detection/', 'Bot detection')}). It refuses PerplexityBot there too by default. If Perplexity sends you readers, let PerplexityBot in; it then reads those pages free.` },
+];
+
+const TODO = [
+ { title: 'Ask whether Comet Plus still pays.', text: 'It isn\'t on Perplexity\'s pricing page or plan guide. Ask before you plan on it.' },
+ { title: 'Ask how your share is computed.', text: 'Ask for your counts of human visits, search citations and agent actions each period, the weight on each, and the pool total. Without them you can\'t recompute your share.' },
+ { title: 'Ask what the pool holds.', text: `Digiday reported that Pro and Max revenue is pooled; Press Gazette reported "$5 from the more expensive Pro and Max subscriptions" (${a(X.pressGazette, 'Press Gazette')}, published 2025-10-02). Ask which.` },
+ { title: 'Ask how a subscriber gets past your paywall.', text: 'Perplexity doesn\'t document it, so ask before your team estimates the work.' },
+ { title: 'Ask how Comet\'s visits show in your numbers.', text: 'Perplexity doesn\'t document which user agent Comet sends. Ask whether its visits count as human pageviews in your analytics and ad reporting, and how to tell them apart. Ask your ad partners too.' },
+ { title: 'Read the contract for limits and exclusivity.', text: 'See For legal.' },
+ { title: 'Set your crawler rules.', text: 'Decide whether PerplexityBot reads your licensed pages, which FORA\'s edge refuses by default (see Search traffic). Keep Perplexity\'s published IP ranges current in your firewall.' },
+ { title: 'Sell to other agents as well.', text: `Name your FORA Exchanges in fora.json and set your price per resource. FORA itself is free, open-source software. For the publisher console's fees, see About FORA.` },
+];
+
+const LEGAL = [
+ { title: 'No standard agreement.', text: 'Joining is an email and a private contract. Perplexity publishes no standard agreement.' },
+ { title: 'Use in answers.', text: `CNN's complaint names "limits on Perplexity's use of CNN content in its answers to users" among the issues CNN and Perplexity couldn't agree on (${a(X.cnn, 'CNN\'s complaint')}, ¶47). Settle how Perplexity may use your content in answers.` },
+ { title: 'Exclusivity.', text: 'Whether a Comet Plus contract limits licensing the same content to other AI companies isn\'t public. Check yours before you set FORA terms for the same uses.' },
+ { title: 'Disputes.', text: `Perplexity publishes no dispute process. Publishers' disagreements with Perplexity have gone to court as copyright suits: CNN, The New York Times and Dow Jones, among others, have sued it ${cite(9, 15, 16)}.` },
+];
+
+const TIMELINE = [
+ { title: 'July 2024: an ad-revenue share.', text: `The Publishers' Program shared ad revenue "when Perplexity earns revenue from an interaction where a publisher's content is referenced" (${a(X.program, 'Publishers\' Program')}). Of that ad share, Chan told Digiday in August 2025: "we are not disclosing any of those numbers." Asked whether checks had gone out to publishers, she said: "Checks have gone out."` },
+ { title: 'From the end of 2025: no ads.', text: `The Financial Times reported, as Gigazine relayed, that Perplexity would phase out its ads from the end of 2025 ${cite(13)}.` },
+ { title: 'October 2025: CNN, a launch partner.', text: `CNN's deal began as a term sheet, "with an agreement to finalize the contract within sixty days" (¶46) ${cite(15)}. Perplexity's partner page listed CNN as a launch partner from 2025-10-02 until at least 2025-12-04 (${a(X.archive, 'Internet Archive, capture of 2025-12-04')}).` },
+ { title: 'November 2025: the term sheet ends.', text: `CNN's complaint says the term sheet never became a contract, over issues that include limits on use in answers (see For legal). It says the two agreed to end the term sheet on 2025-11-24 (¶47) ${cite(15)}.` },
+ { title: 'May 2026: CNN sues.', text: `CNN sued Perplexity on 2026-05-28. Perplexity's response, to TheWrap: "You can’t copyright facts." ${cite(19)}` },
+ { title: 'November 2026: off the pricing page.', text: `Perplexity's ${a(X.pricing, 'pricing page')} and ${a(X.plans, 'plan guide')} don't list Comet Plus, and Perplexity hasn't announced its end.` },
+];
+
+// The SourceList, numbered in order of first citation: an entry's position,
+// counted across the groups, is the N in cite(N).
+const SOURCES = [
+ {
+ name: 'Perplexity',
+ items: [
+ { title: 'Introducing Comet Plus', publisher: 'Perplexity', href: X.comet, checked: '2026-11-15' },
+ { title: 'Announcing Comet Plus launch partners', publisher: 'Perplexity', href: X.partners, checked: '2026-11-14' },
+ { title: 'Introducing the Perplexity Publishers\' Program', publisher: 'Perplexity', href: X.program, checked: '2026-11-14' },
+ { title: 'Pricing', publisher: 'Perplexity', href: X.pricing, checked: '2026-11-15' },
+ { title: 'Perplexity Crawlers', publisher: 'Perplexity', href: X.crawlers, checked: '2026-11-15' },
+ { title: 'How does Perplexity follow robots.txt?', publisher: 'Perplexity', href: X.robots, checked: '2026-11-15' },
+ { title: 'Bullying is not innovation', publisher: 'Perplexity', href: X.bullying, checked: '2026-11-14' },
+ { title: 'Signing-key directory address (returns HTTP 404)', publisher: 'Perplexity', href: X.keys, checked: '2026-11-15' },
+ { title: 'About the Dow Jones lawsuit', publisher: 'Perplexity', href: X.dowJones, checked: '2026-11-14' },
+ { title: 'API Terms of Service - Search Service', publisher: 'Perplexity', href: X.searchApi, checked: '2026-11-14' },
+ { title: 'Which Perplexity Subscription Plan is right for you?', publisher: 'Perplexity', href: X.plans, checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Press, filings and others',
+ items: [
+ { title: 'How Perplexity\'s new revenue model works, according to its head of publisher partnerships', publisher: 'Digiday', href: X.digiday, checked: '2026-11-14' },
+ { title: 'Perplexity will phase out ads introduced in 2024 from the end of 2025', publisher: 'Gigazine', href: X.gigazine, checked: '2026-11-14' },
+ { title: 'Media Briefing: Here are the hurdles to Perplexity\'s pitch as the publisher-friendly LLM', publisher: 'Digiday', href: X.hurdles, checked: '2026-11-14' },
+ { title: 'Cable News Network, Inc. v. Perplexity AI, Inc., complaint', publisher: 'U.S. District Court for the Southern District of New York', href: X.cnn, checked: '2026-11-14' },
+ { title: 'Form 10-Q for the quarter ended September 30, 2026', publisher: 'The New York Times Company', href: X.nyt, checked: '2026-11-14' },
+ { title: 'Perplexity launches AI subscription revenue-share scheme for publishers', publisher: 'Press Gazette', href: X.pressGazette, checked: '2026-11-14' },
+ { title: 'Announcing Comet Plus launch partners, capture of 2025-12-04', publisher: 'Internet Archive', href: X.archive, checked: '2026-11-14' },
+ { title: 'CNN Sues Perplexity for Unlawfully Distributing News Content, AI Company Responds: ‘You Can’t Copyright Facts’', publisher: 'TheWrap', href: X.thewrap, checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'FORA',
+ items: [
+ { title: 'Publisher onboarding', publisher: 'FORA documentation', href: '/getting-started/publisher-onboarding/', checked: '2026-11-15' },
+ { title: 'What is FORA?', publisher: 'FORA documentation', href: '/getting-started/what-is-fora/', checked: '2026-11-15' },
+ { title: 'Bot detection', publisher: 'FORA documentation', href: '/components/edge-function/bot-detection/', checked: '2026-11-15' },
+ { title: 'Licensing terms', publisher: 'FORA documentation', href: '/protocol/licensing-terms/', checked: '2026-11-15' },
+ { title: 'Broker', publisher: 'FORA documentation', href: '/components/broker/overview/', checked: '2026-11-15' },
+ { title: 'Authentication', publisher: 'FORA documentation', href: '/protocol/authentication/', checked: '2026-11-15' },
+ ],
+ },
+];
+---
+
+
+ Where Perplexity is ahead
+ Run both
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
PerplexityBot builds Perplexity's search index, and it follows robots.txt. How refusing it changes your Comet Plus share isn't public.
+
+
+
+
+
+
+
+
+
+
+
Both are ways to license Perplexity's use of your content, and they compose on your site, where you choose what each Perplexity crawler gets.
+
+
+
+
Works today
+
+
Doesn't fit
+
Comet Plus as a marketplace for FORA agents. A FORA Broker presents offers that a seller signs. Perplexity licenses no partner's article to other AI companies. Its Search API sells results that its terms say are "sourced from the public internet", and the customer doesn't own them (${a(X.searchApi, 'Search API terms')}). There is no signed license offer to present.`} />
+
Paying FORA sales through Comet Plus. Perplexity offers no payment service to anyone else. It allocates the pool from its own traffic counts.'} />
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Sources are checked on the date of publication.
+
+
+
+
+
+
+
+
+
+ What is FORA?
+ Run your own marketplace
+ Licensing terms
+ All content marketplaces
+
+
+
+
diff --git a/website/src/pages/resources/content-marketplaces/prorata-gist.astro b/website/src/pages/resources/content-marketplaces/prorata-gist.astro
new file mode 100644
index 00000000..51e3e3bc
--- /dev/null
+++ b/website/src/pages/resources/content-marketplaces/prorata-gist.astro
@@ -0,0 +1,369 @@
+---
+// Resource: ProRata.ai and Gist, and FORA, one of the deep pages of the
+// content-marketplaces series (hub: /resources/content-marketplaces/). Laid
+// out in the vendor page template, as the TollBit page is: hero, at a glance,
+// the short answer, how you're paid, key facts, how your articles get into
+// answers, who has been paid, what Gist Answers needs from your site, what can
+// go wrong, where ProRata is ahead, the eight questions, running both, what to
+// do, and About FORA with the sources and the series navigation.
+//
+// Shared texts (About FORA, the definitions, the search callout, the table
+// questions, the GlanceCard rows and FORA's fixed cells) come from
+// src/data/content-marketplaces.ts. A claim cites its source inline, as a
+// link, or with cite(N), which links to entry N of the numbered SourceList at
+// the foot; SOURCES below is numbered in order of first citation, so an
+// entry's position is its number. Text with links is HTML and goes through
+// set:html.
+import LandingLayout from '../../../layouts/LandingLayout.astro';
+import AudienceHero from '../../../components/landing/AudienceHero.astro';
+import Band from '../../../components/landing/Band.astro';
+import GlanceCard from '../../../components/landing/GlanceCard.astro';
+import CardGrid from '../../../components/landing/CardGrid.astro';
+import Callout from '../../../components/landing/Callout.astro';
+import StepList from '../../../components/landing/StepList.astro';
+import Figure from '../../../components/landing/Figure.astro';
+import KeyFacts from '../../../components/landing/KeyFacts.astro';
+import PullQuote from '../../../components/landing/PullQuote.astro';
+import PointList from '../../../components/landing/PointList.astro';
+import CompareTable from '../../../components/landing/CompareTable.astro';
+import StatusTag from '../../../components/landing/StatusTag.astro';
+import SourceList from '../../../components/landing/SourceList.astro';
+import SeriesNav from '../../../components/landing/SeriesNav.astro';
+import ClosingCta from '../../../components/landing/ClosingCta.astro';
+import { resourceUpdated } from '../../../data/resources';
+import {
+ ABOUT_FORA, DEFINITIONS, SEARCH_FORA, FORA_CELLS, SEE_ABOUT, compareRows, glanceGroups,
+} from '../../../data/content-marketplaces';
+import { cite, statusTag } from '../../../lib/resource-markup';
+
+const HREF = '/resources/content-marketplaces/prorata-gist/';
+const IMG = '/resources/content-marketplaces/prorata-gist';
+
+// Every external source the page links.
+const P = {
+ msa: 'https://gist.ai/support/ga/wpnewpartner',
+ ads: 'https://gist.ai/gist-ads-discontinued',
+ guide: 'https://gist.ai/support/ga/wpuserguide',
+ answers: 'https://gist.ai/products/gist-answers',
+ products: 'https://prorata.ai/products',
+ how: 'https://prorata.ai/how-it-works',
+ paper: 'https://prorata.ai/resources/without-attribution-quality-content-cannot-survive',
+ terms: 'https://gist.ai/terms-of-use',
+ nma: 'https://www.newsmediaalliance.org/prorata-licensing-partnership/',
+ dpcmo: 'https://dpcmo.dk/2025/12/10/prorata-partners-with-dpcmo-to-launch-the-first-decentralized-sovereign-ai-answer-engine/',
+};
+const a = (href: string, text: string) => `${text}`;
+
+const GLANCE = glanceGroups({
+ payer: [`ProRata itself, monthly, by bank transfer once you're owed more than $1,000. Its Gist Ads network ended on 2026-09-01; Gist Answers still runs. ProRata hasn't said what funds the pool since ${cite(1, 2)}.`, FORA_CELLS.payer],
+ fees: [`A share of a pool, not a rate: see How you're paid ${cite(1)}.`, FORA_CELLS.fees],
+ reporting: [
+ `A monthly statement, and per article: answers it was used in, its average attribution, citation clicks. Not the month's total revenue or answer count ${cite(1, 3)}.`,
+ `Each sale in your operator's console: gross, the Exchange fee and net, and a check of both signatures ${cite(9)}.`,
+ ],
+ contract: [
+ `ProRata's online MSA, accepted on a web form. Non-exclusive. ProRata can change it; if you object, you can only end it ${cite(1)}.`,
+ `Your agreement with your Exchange operator. Your terms go into each offer the Exchange signs ${cite(10)}.`,
+ ],
+ status: [
+ `${statusTag('live')} "150+" publishing partners, by ProRata's count. No payout total published ${cite(4)}.`,
+ `${statusTag('live')} ${SEE_ABOUT}`,
+ ],
+ dns: [`No change ${cite(3)}.`, `No record to add ${cite(9)}.`],
+ cdn: [
+ `No code. Cache rules must keep the feed working, and you purge the cache after a correction ${cite(3)}.`,
+ 'A drop-in edge function. The rest depends on your own setup; FORA adds to what you run rather than replacing it.',
+ ],
+ firewall: [
+ `An allow rule for ProRata's crawler. Its user agent and IP ranges aren't published: ask ProRata ${cite(3)}.`,
+ `Leave your licensed paths and your discovery file to the edge function, or keep your bot product on and let signed requests and the file through ${cite(11)}.`,
+ ],
+ site: [
+ `The WordPress plugin or one script tag. Posts only, by default. How other CMSs get in isn't documented ${cite(3, 4)}.`,
+ `One discovery file on your site, and a sitemap the operator imports your pages from ${cite(9)}. No WordPress plugin.`,
+ ],
+ search: [
+ `Gist blocks no bot ${cite(3)}.`,
+ `Googlebot, Bingbot and Applebot always pass ${cite(11)}. See Search traffic, below.`,
+ ],
+ upkeep: [
+ `WordPress's scheduler must run, or nothing is sent. Removals go by email ${cite(3)}.`,
+ `The operator imports your pages on a schedule. Editing your discovery file needs no redeploy ${cite(9)}.`,
+ ],
+ leaving: [`End the MSA. Content already ingested stays until you email the URLs ${cite(1, 3)}.`, FORA_CELLS.leaving],
+});
+
+const SHORT_ANSWER = [
+ { title: 'ProRata measures the answer.', text: `Its engine splits each Gist answer into claims, finds the sources that support them, and divides a pool of half its net revenue by those shares.` },
+ { title: 'FORA prices the access.', text: 'You set the price, the unit and the terms for each resource, and your Exchange signs each offer. Any agent your terms allow can buy on them, and each purchase is a signed record.' },
+ { title: 'They fit on one site.', text: 'ProRata\'s contract is non-exclusive, so you can sell through FORA on the same site. Gist needs its WordPress plugin or a script tag.' },
+];
+
+const PAID_STEPS = [
+ { title: 'Half the month\'s revenue becomes a pool.', text: `Half of ProRata's net revenue from its answer products goes to "the pool of all content partners" (${a(P.msa, 'MSA §5')}). That revenue comes from "advertising, subscriptions, and sponsorships (if any)" in ProRata's own products.` },
+ { title: 'Every answer that month is worth the same.', text: 'Each valid answer, as ProRata counts them, carries the month\'s revenue divided by the month\'s valid answers. So what one answer earned ProRata doesn\'t change what it pays you.' },
+ { title: 'Your share depends on how much of the answer is yours.', text: `You get that amount, times 50%, times your attribution percent: the share of the answer's claims your content supplied (${a(P.msa, 'MSA §5')}).` },
+ { title: 'You\'re paid above $1,000.', text: `ProRata sends a monthly statement and pays by bank transfer "once the amount due to Partner exceeds one thousand dollars ($1,000)" (${a(P.msa, 'MSA §5')}). Until then, your earnings carry over.` },
+];
+
+const KEY_FACTS = [
+ { value: '50%', text: `of ProRata's net revenue from its answer products goes to the partner pool ${cite(1)}` },
+ { value: '50%', text: `of the widget's ad revenue on your site is yours ${cite(1)}` },
+ { value: '1 year', text: `ProRata keeps the records you can have inspected ${cite(1)}` },
+ { value: '7 days', text: `"from kickoff to live for a typical site, depending on your CMS and catalog size", by ProRata's count ${cite(4)}` },
+];
+
+const ARTICLE_STEPS = [
+ { title: 'You sign.', text: `You accept ProRata's Online Master Services Agreement (MSA) on a web form, and an API key follows "within 2 business days upon approval" (${a(P.msa, 'sign-up page and MSA')}).` },
+ { title: 'ProRata reads your site.', text: `You paste one script tag or install the WordPress plugin. The plugin registers your public posts feed and sends each post's address when the post is saved. ProRata's crawler, a program that fetches your pages, then reads the post through the feed: "Gist Answers sees only what an anonymous visitor sees" (${a(P.guide, 'WordPress user guide')}).` },
+ { title: 'Readers ask.', text: `The Gist widget on your pages, and ProRata's own answer site, ask.gist.ai, answer readers' questions. You choose the sources: your site, your whole organization, or the "Full Gist network" (${a(P.answers, 'Gist Answers')}).` },
+ { title: 'The engine measures each answer.', text: `It breaks the answer into claims, "atomic units of fact", and matches them to the sources. Then it computes "the proportion of the answer's total claims provided by each source and creator" (${a(P.how, 'How it works')}). The white paper names three weights, for originality, specificity and commonality, and gives no values for them (${a(P.paper, 'white paper')}).` },
+];
+
+const ENGINEERS = [
+ { title: 'Content.', text: 'By default ProRata takes posts only, through your public posts feed, /wp-json/wp/v2/posts. Pages and custom post types need a filter in code, and "Adding a post type with a filter is not retroactive".' },
+ { title: 'Sending.', text: 'The plugin queues each post\'s address on save, and WordPress\'s scheduler sends the queue.' },
+ { title: 'Cache.', text: 'The feed must honor its date filter, which "a page or edge cache rule, a security rule, or a plugin that changes REST queries" can break. The plugin "does not bypass or purge any cache", so after a correction you purge the post and the feed yourself.' },
+ { title: 'Bot protection.', text: 'Your bot protection has to let ProRata\'s crawler read the feed: a 401 or 403 "means the crawl of the feed fails". The crawler\'s user agent and IP ranges aren\'t published, so an allow rule needs them from ProRata.' },
+ { title: 'Widget and logs.', text: 'Readers\' browsers load the widget from Gist\'s own CDN. There is no log stream; the console shows per-article numbers and the MSA adds the monthly statement.' },
+];
+
+const CAUTIONS = [
+ { title: 'Paywalled text arrives empty.', text: 'A password-protected post, or text a paywall plugin hides, reaches Gist empty: "Gist Answers can answer from neither".' },
+ { title: 'A stopped scheduler stops everything.', text: 'Sending runs on WordPress\'s scheduler: "Without it, queued content is never sent."' },
+ { title: 'No consent control.', text: 'For consent banners, "There is no consent setting and no JavaScript API to load or unload the widget".' },
+];
+
+// Six points; the edge function's define callout follows the fourth, the
+// point that first names it.
+const AHEAD_1 = [
+ { title: 'It measures how much of an answer you supplied.', text: `ProRata returns "attribution scorecards at claim, article, publication, and publisher levels" (${a(P.how, 'How it works')}). ProRata can pay you by your share of each answer; FORA cannot. A FORA price is fixed when the agent buys, and a usage report, the buyer's later account of its use, "never re-prices a purchase" (${a('/protocol/licensing-terms/#metering-modes', 'Licensing terms')}).` },
+ { title: 'It publishes its payout rule.', text: `The pool, the split per answer and the $1,000 threshold are in a contract anyone can read (${a(P.msa, 'MSA §5')}). The console shows each article's "Average attribution", "the percentage of an answer that an article's content makes up" (${a(P.guide, 'user guide')}). The attribution itself is ProRata's to compute: the weights' values aren't published. FORA is software and sets no payout terms: each Exchange operator pays you under its own agreement (${a('/getting-started/what-is-fora/', 'What is FORA?')}).` },
+ { title: 'It pays you itself, and gives your readers a product.', text: `You don't wait for an AI company to buy: ProRata pays the pool from its own answer products. The widget answers readers on your pages at no AI cost to you, and the MSA gives you half of the widget's ad revenue on your site (${a(P.msa, 'MSA §5')}). ProRata says readers who use Gist Answers on one McClatchy site, Woman's World, stay about three times as long. That's its own data, with no period given (${a(P.answers, 'Gist Answers')}). On FORA, any agent your terms allow buys from your Exchange, and your operator pays you.` },
+ { title: 'It puts no code on your CDN.', text: `Gist needs one script tag or plugin on your pages, and its WordPress plugin "handles ingestion, configuration, management and reporting" (${a(P.answers, 'Gist Answers')}). Your work is the feed, your cache rules and an allow rule for the crawler, whose identity you get from ProRata. FORA asks more: you set your terms in an Exchange operator's console, serve one file that names your Exchanges, and add FORA's edge function to your CDN (${a('/getting-started/publisher-onboarding/', 'Publisher onboarding')}). A WordPress plugin for FORA is planned ${statusTag('roadmap')}.` },
+];
+const AHEAD_2 = [
+ { title: 'It gives you an audit right.', text: `If you dispute a payment, an independent third party can inspect ProRata's records, kept for a year after each payment (${a(P.msa, 'MSA §6')}). FORA's protocol gives a publisher no audit right. ${SEE_ABOUT}` },
+ { title: 'It reaches you through your trade association.', text: `The News/Media Alliance, which by its own count represents "more than 2,200" news, magazine and digital media organizations, lets its members opt into a license with ProRata (${a(P.nma, 'N/MA')}, announced 2025-03-26). Denmark's DPCMO offers its members "a pre-negotiated partnership agreement" (${a(P.dpcmo, 'DPCMO')}, announced 2025-12-10). With FORA, an association can run an Exchange for its members.` },
+];
+
+const COMPARE = compareRows([
+ [`You click to accept the MSA, which grants fixed uses: answers in ProRata's products, and attribution. Nothing on your site says so to anyone else ${cite(1)}.`, `Your fora.json names the Exchanges that sell your content; you set terms per resource in the Exchange operator's console ${cite(9, 10)}.`],
+ [`No outside buyer does. ProRata's engine picks sources from its own index of partner content and doesn't publish how it picks them. On your widget, "Answers come from your content first" ${cite(4)}.`, `From your Exchange, or through a Broker (${a('/resources/content-marketplaces/#broker-order', 'how a Broker orders offers')}) ${cite(13)}.`],
+ [`There is no outside buyer. ProRata's crawler publishes no user agent or IP ranges ${cite(3)}.`, `Each request carries a signature, checked against a key the buyer publishes on its own domain (Web Bot Auth, the IETF working group's -00 draft for signed bot requests) ${cite(10)}.`],
+ [`The MSA grant, for every answer ProRata generates during the contract ${cite(1)}.`, `An offer the Exchange signs and the agent accepts, and a signed link that expires, checked at your CDN ${cite(10)}.`],
+ [`ProRata: 50% of net revenue, split per valid answer by attribution. The online MSA sets the same split for every partner who accepts it. ProRata can change the terms ${cite(1)}.`, `You: free, per unit (pages, text tokens, seconds or your own) or a flat fee, signed on each offer. No revenue-share price, by design: a share is a free term shown to agents that state the deal's scope, with a reporting duty (${a('/protocol/licensing-terms/#revenue-share-arrangements', 'Revenue share')}).`],
+ [`ProRata pays by bank transfer, monthly, once you're owed more than $1,000 ${cite(1)}.`, FORA_CELLS.money],
+ [`A monthly statement per partner, publication and article (above a minimum number of answers): answers, average attribution, revenue. Not the month's total revenue or answer count ${cite(1)}.`, `The operator's console: each sale with gross, fee and net, and a check of both signatures. The console is each operator's own; no protocol call returns your sales ${cite(9)}.`],
+ [`An independent inspection of one year of records, then the courts of Los Angeles County ${cite(1)}.`, FORA_CELLS.disputes],
+]);
+
+const ONE_SITE = [
+ { title: 'Run both.', text: `ProRata's licenses "are nonexclusive" (${a(P.msa, 'MSA §13')}), and Gist Answers "fits alongside anything else in your AI strategy" (${a(P.answers, 'Gist Answers')}). Your FORA discovery file names your FORA Exchanges; the ProRata deal stays a private contract.` },
+ { title: 'Let ProRata\'s crawler through your bot product.', text: `On most CDNs the bot product runs before FORA's edge function, so it can block the crawler first (${a('/components/edge-function/bot-detection/', 'Bot detection')}). Allow the crawler there, with the user agent and IP ranges you get from ProRata. FORA's edge function recognizes AI bots mainly by the name they give, and ProRata's crawler isn't on its list, so the edge lets it through (${a('/getting-started/publisher-onboarding/', 'Publisher onboarding')}).` },
+ { title: 'If you license your articles at the edge, license the feed too.', text: `Your WordPress posts feed serves the same article text to any client. If only the article pages are licensed, an AI bot can read the text from the feed instead. So license the feed's path with your articles (${a('/getting-started/publisher-onboarding/', 'Publisher onboarding')}). Before you do, ask your engineers what else reads the feed, such as your mobile app or your AMP pages, and test them after the change.` },
+];
+
+const TODO = [
+ { title: 'Ask for the pool totals.', text: 'Your statement doesn\'t show the month\'s total revenue or total valid answers, so you can\'t recompute your share. Ask for both, which answers count as valid, and where the share of each answer that no partner gets goes.' },
+ { title: 'Ask what funds the pool.', text: 'Ask which products pay into the pool now that Gist Ads has ended, and whether the widget on your pages still carries ads, and whose.' },
+ { title: 'Check the feed as a stranger would.', text: 'Open your posts feed in a private browser window, with and without a date filter, as the user guide shows. Ask support@gist.ai for the crawler\'s user agent and IP ranges, and allow it at your CDN.' },
+ { title: 'Keep your own removal list.', text: `The plugin sends no removal signal, so content stays in Gist until you email the URLs (${a(P.guide, 'user guide')}). Ask what happens to the content already ingested if you end the MSA.` },
+ { title: 'Audit within the year.', text: `ProRata keeps its records for one year after each payment (${a(P.msa, 'MSA §6')}).` },
+ { title: 'Sell to agents as well.', text: `ProRata's contract is non-exclusive. Choose an Exchange operator (see About FORA), name your FORA Exchanges in fora.json, set your price per resource, and cover the posts feed at the edge.` },
+];
+
+const LEGAL = [
+ { title: 'ProRata can change the terms.', text: `It amends the MSA by posting a new version. "If you object to any such changes, your sole recourse will be to terminate this Agreement" (${a(P.msa, 'MSA §13')}).` },
+ { title: 'Non-exclusive.', text: '"The rights and licenses that each Party grants to the other Party under this Agreement are nonexclusive" (MSA §13).' },
+ { title: 'Audit.', text: 'A partner that "reasonably disputes Revenue Share paid" can have "an independent third party confidentially inspect the Records", kept for a year after each payment (MSA §6).' },
+ { title: 'Removal.', text: 'After written notice, ProRata stops using the content "within one (1) business day" (MSA §7). The plugin sends no removal signal, so the notice is yours to send.' },
+ { title: 'Consent.', text: 'The widget has no consent setting (see What can go wrong).' },
+ { title: 'Venue.', text: '"The federal and state courts located in Los Angeles County will have sole and exclusive jurisdiction" (MSA §13).' },
+ { title: 'Version.', text: 'The MSA quoted here is the one "Last updated: October 27, 2026".' },
+];
+
+// The SourceList, numbered in order of first citation: an entry's position,
+// counted across the groups, is the N in cite(N).
+const SOURCES = [
+ {
+ name: 'ProRata.ai and Gist',
+ items: [
+ { title: 'ProRata Online Master Services Agreement (MSA)', publisher: 'ProRata.ai', href: P.msa, checked: '2026-11-15' },
+ { title: 'Gist Ads discontinued', publisher: 'ProRata.ai', href: P.ads, checked: '2026-11-14' },
+ { title: 'Gist Answers for WordPress: user guide', publisher: 'ProRata.ai', href: P.guide, checked: '2026-11-15' },
+ { title: 'Gist Answers', publisher: 'ProRata.ai', href: P.answers, checked: '2026-11-14' },
+ { title: 'Products', publisher: 'ProRata.ai', href: P.products, checked: '2026-11-14' },
+ { title: 'How it works', publisher: 'ProRata.ai', href: P.how, checked: '2026-11-14' },
+ { title: 'Without attribution, quality content cannot survive (white paper)', publisher: 'ProRata.ai', href: P.paper, checked: '2026-11-14' },
+ { title: 'Terms of Use', publisher: 'Gist', href: P.terms, checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'FORA',
+ items: [
+ { title: 'Publisher onboarding', publisher: 'FORA documentation', href: '/getting-started/publisher-onboarding/', checked: '2026-11-15' },
+ { title: 'What is FORA?', publisher: 'FORA documentation', href: '/getting-started/what-is-fora/', checked: '2026-11-15' },
+ { title: 'Bot detection', publisher: 'FORA documentation', href: '/components/edge-function/bot-detection/', checked: '2026-11-15' },
+ { title: 'Licensing terms', publisher: 'FORA documentation', href: '/protocol/licensing-terms/', checked: '2026-11-15' },
+ { title: 'Broker', publisher: 'FORA documentation', href: '/components/broker/overview/', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Press, filings and others',
+ items: [
+ { title: 'ProRata licensing partnership', publisher: 'News/Media Alliance', href: P.nma, checked: '2026-11-14' },
+ { title: 'ProRata partners with DPCMO', publisher: 'DPCMO', href: P.dpcmo, checked: '2026-11-14' },
+ ],
+ },
+];
+---
+
+
+ Where ProRata is ahead
+ Run both
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Gist doesn't block, meter, redirect or charge any bot, so every bot rule on your site stays yours, and search crawlers reach your site as today.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
On one site
+
+
Doesn't fit
+
Gist answers as FORA offers. Gist sells agents nothing: its terms forbid access by any "agent or other automatic device" (${a(P.terms, 'Terms of Use')}). A Gist answer is also written when someone asks, and FORA sells resources that exist before the sale (${a('/getting-started/what-is-fora/#scope', 'What is FORA?')}). ProRata offers no payment service to others either, so an Exchange can't settle through it.`} />
+
+
+
+
+
+
+
+
+
+
+
+
+
Sources
+
Sources are checked on the date of publication.
+
+
+
+
+
+
+ Licensing terms
+ Run your own marketplace
+ All content marketplaces
+
+
+
+
diff --git a/website/src/pages/resources/content-marketplaces/tollbit.astro b/website/src/pages/resources/content-marketplaces/tollbit.astro
new file mode 100644
index 00000000..e3d07e91
--- /dev/null
+++ b/website/src/pages/resources/content-marketplaces/tollbit.astro
@@ -0,0 +1,406 @@
+---
+// Resource: TollBit and FORA, one of the deep pages of the content-marketplaces
+// series (hub: /resources/content-marketplaces/). Laid out in the vendor page
+// template: hero, at a glance, the short answer, how you're paid, key facts,
+// who has been paid, what it takes on your site, what can go wrong, where
+// TollBit is ahead, the eight questions, running both, what to do, and About
+// FORA with the sources and the series navigation.
+//
+// Shared texts (About FORA, the definitions, the search callout, the table
+// questions, the GlanceCard rows and FORA's fixed cells) come from
+// src/data/content-marketplaces.ts. A claim cites its source inline, as a
+// link, or with cite(N), which links to entry N of the numbered SourceList at
+// the foot; SOURCES below is numbered in order of first citation, so an
+// entry's position is its number. Text with links is HTML and goes through
+// set:html.
+import LandingLayout from '../../../layouts/LandingLayout.astro';
+import AudienceHero from '../../../components/landing/AudienceHero.astro';
+import Band from '../../../components/landing/Band.astro';
+import GlanceCard from '../../../components/landing/GlanceCard.astro';
+import CardGrid from '../../../components/landing/CardGrid.astro';
+import Callout from '../../../components/landing/Callout.astro';
+import StepList from '../../../components/landing/StepList.astro';
+import Figure from '../../../components/landing/Figure.astro';
+import KeyFacts from '../../../components/landing/KeyFacts.astro';
+import PullQuote from '../../../components/landing/PullQuote.astro';
+import Split from '../../../components/landing/Split.astro';
+import PointList from '../../../components/landing/PointList.astro';
+import CompareTable from '../../../components/landing/CompareTable.astro';
+import StatusTag from '../../../components/landing/StatusTag.astro';
+import SourceList from '../../../components/landing/SourceList.astro';
+import SeriesNav from '../../../components/landing/SeriesNav.astro';
+import ClosingCta from '../../../components/landing/ClosingCta.astro';
+import { resourceUpdated } from '../../../data/resources';
+import {
+ ABOUT_FORA, DEFINITIONS, SEARCH_FORA, FORA_CELLS, SEE_ABOUT, compareRows, glanceGroups,
+} from '../../../data/content-marketplaces';
+import { cite, statusTag } from '../../../lib/resource-markup';
+
+const HREF = '/resources/content-marketplaces/tollbit/';
+const IMG = '/resources/content-marketplaces/tollbit';
+
+// Every external source the page links.
+const T = {
+ overview: 'https://tollbit.com/docs/',
+ tokens: 'https://docs.tollbit.com/docs/tokens',
+ settingRates: 'https://docs.tollbit.com/docs/setting-rates',
+ security: 'https://docs.tollbit.com/docs/security',
+ terms: 'https://tollbit.com/legal/publisher-platform-agreement/',
+ license: 'https://gateway.tollbit.com/license/j4ikxb8qqsu2qu91q78a9e70/ON_DEMAND_LICENSE_ka02vqvjacoy2i5vs8z90hxm',
+ verifying: 'https://docs.tollbit.com/docs/verifying-a-property',
+ cloudflare: 'https://docs.tollbit.com/docs/cloudflare',
+ integrations: 'https://docs.tollbit.com/docs/integrations',
+ botManagement: 'https://docs.tollbit.com/docs/bot-management',
+ licensedSearch: 'https://docs.tollbit.com/docs/licensed-search',
+ rates: 'https://docs.tollbit.com/docs/rates',
+ akamai: 'https://tollbit.com/blog/akamai-partnership/',
+ seriesA: 'https://tollbit.com/blog/series-a/',
+ home: 'https://tollbit.com/',
+ staticIps: 'https://tollbit.com/static-ips.txt',
+ architecture: 'https://docs.tollbit.com/docs/platform-architecture',
+ other: 'https://docs.tollbit.com/docs/other',
+ configure: 'https://docs.tollbit.com/docs/step-3-configure-agent-site',
+ cache: 'https://docs.tollbit.com/docs/content-cache',
+ defaults: 'https://docs.tollbit.com/docs/content-formatting-1',
+ aws: 'https://docs.tollbit.com/docs/aws-s3-cloudfront-alb',
+ devOverview: 'https://docs.tollbit.com/docs/dev-overview',
+ analytics: 'https://docs.tollbit.com/docs/step-2-enable-analytics',
+ content: 'https://docs.tollbit.com/docs/content',
+ edgeConfig: 'https://github.com/FORA-Protocol/reference-implementation/blob/main/src/edge/CONFIGURATION.md',
+ tnl: 'https://www.prnewswire.com/news-releases/tnl-mediagene-announces-early-success-in-ai-content-licensing-revenue-model-via-tollbit-marketplace-integration-and-strategic-partnership-302647966.html',
+ digiday: 'https://digiday.com/media/the-washington-posts-arc-xp-adds-tollbit-to-help-publishers-make-money-from-ai-bot-traffic/',
+};
+const a = (href: string, text: string) => `${text}`;
+
+const GLANCE = glanceGroups({
+ payer: [`TollBit, through Stripe, early each month, in US dollars ${cite(1, 2)}.`, FORA_CELLS.payer],
+ fees: [`You keep your full rate. The buyer pays a fee on top; its size isn't published ${cite(3)}.`, FORA_CELLS.fees],
+ reporting: [
+ `Daily bot analytics, and each sale: agent, page, price ${cite(3, 4)}.`,
+ `Each sale in your operator's publisher console, with totals and a daily CSV. No bot-traffic report ${cite(26)}.`,
+ ],
+ contract: [
+ `A license between you and each buyer; TollBit isn't a party ${cite(5)}. New York law ${cite(6)}. No dispute call.`,
+ `Your agreement with your Exchange operator. Your terms go into each offer the Exchange signs ${cite(26, 27)}. Disputes: see About FORA.`,
+ ],
+ status: [
+ `${statusTag('live')} A publisher reports revenue ${cite(36)}. No TollBit page names a big AI company as a buyer.`,
+ `${statusTag('live')} ${SEE_ABOUT}`,
+ ],
+ dns: [`A TXT record, and an NS record that hands a subdomain to TollBit ${cite(7)}.`, `No record to add ${cite(26)}.`],
+ cdn: [
+ `One forwarding rule for 69 AI user agents ${cite(8)}. Guides for 12 platforms ${cite(9)}.`,
+ 'A drop-in edge function. The rest depends on your own setup; FORA adds to what you run rather than replacing it.',
+ ],
+ firewall: [
+ `An allow rule for TollBit's fetcher in every layer ${cite(7)}.`,
+ `On most CDNs they run first, so they must let signed requests and your two licensing files through. Keep them: the edge function knows AI agents mainly by the name they send ${cite(26, 28)}.`,
+ ],
+ site: [
+ `No change. TollBit fetches, cleans and serves the page. Your HTTP logs, readers' requests included, go to TollBit ${cite(1, 4)}.`,
+ `Serve one file, fora.json, from your site. The operator imports your pages from your sitemap ${cite(26)}.`,
+ ],
+ search: [
+ `Googlebot and Bingbot aren't on TollBit's list; they reach your site as today ${cite(8)}.`,
+ `Googlebot, Bingbot and Applebot always pass ${cite(28)}. See Search traffic, below.`,
+ ],
+ upkeep: [
+ `On most platforms, update the bot list by hand each quarter ${cite(10)}.`,
+ `Change your Exchanges by editing your FORA file, with nothing to redeploy ${cite(29)}.`,
+ ],
+ leaving: [`No offboarding page. You can end the terms only with no active license ${cite(5)}.`, FORA_CELLS.leaving],
+});
+
+const SHORT_ANSWER = [
+ { title: 'TollBit catches the bots that never asked.', text: 'Your CDN spots AI bots by the name they send and hands them to TollBit, which charges per page at your rate.' },
+ { title: 'FORA serves the agents that do ask.', text: 'An agent signs each request, buys under a signed offer, and fetches with a signed link, which your CDN checks. You price by use, country, buyer type and unit.' },
+ { status: 'design' as const, title: 'TollBit\'s rates behind a FORA Broker.', text: 'A Broker would call TollBit with each agent\'s own TollBit account and show TollBit\'s rates beside your FORA offers.' },
+];
+
+const PAID_STEPS = [
+ { title: 'You set rates per license, per 1,000 pages.', text: `Summarization covers "a summary, grounding, or citation with a single use license". Full Display lets a buyer "display the complete text of an article once". Custom is for a partner you have a deal with. Rates can differ by bot, page, keyword, time and directory (${a(T.settingRates, 'Setting rates')}).` },
+ { title: 'Buyers find you through TollBit and buy a pass for one URL.', text: `One API key searches TollBit's index, which covers sites "across the web" and flags TollBit's own. It shows each URL's licenses and prices (${a(T.licensedSearch, 'Licensed search')}, ${a(T.rates, 'Rates')}).` },
+ { title: 'The pass is short-lived.', text: `TollBit calls it a token: "Tokens expire 5 minutes after issuance." Each works once and names the buyer's agent, the license type and the most the buyer will pay (${a(T.tokens, 'Tokens')}).` },
+ { title: 'TollBit delivers the page.', text: `With a pass, TollBit fetches your page from your site, strips what you exclude, and returns Markdown or HTML (${a(T.security, 'Security')}; ${a(T.overview, 'Platform overview')}, §3.5).` },
+ { title: 'Billing waits for delivery.', text: `Nobody is billed "until the token is actually accepted and a 200 OK response is delivered" (${a(T.overview, 'Platform overview')}, §5.1).` },
+ { title: 'Your rate is your net.', text: `The buyer pays your rate plus "a small transaction fee on top of the rates you set" (${a(T.settingRates, 'Setting rates')}). TollBit doesn't publish the fee's size. You're paid through Stripe early each month, in US dollars (${a(T.overview, 'Platform overview')}, §1.4; ${a(T.tokens, 'Tokens')}).` },
+];
+
+const KEY_FACTS = [
+ { value: '1,000 pages', text: `The unit you set each rate in, per license type ${cite(3)}` },
+ { value: '5 minutes', text: `How long a buyer's pass for one URL lasts. It works once ${cite(2)}` },
+ { value: 'Nearly 20%', text: `Of TollBit's nearly 7,000 sites had earned money by April 2026. The CEO's figure, unaudited ${cite(37)}` },
+ { value: '37 million', text: `"content transactions" TollBit handled in Q2 2025, by its own count. It doesn't say how many were paid ${cite(13)}` },
+];
+
+const SETUP_STEPS = [
+ { title: 'Hand TollBit a subdomain.', text: `A TXT record proves you own the domain. An NS record delegates tollbit.news.example.com to TollBit's nameservers, so TollBit answers for that host under your name (${a(T.verifying, 'Verifying a property')}). TollBit's paywall for bots runs there, as your "Agent Site".` },
+ { title: 'Let TollBit\'s fetcher through.', text: `TollBit pulls your pages from your public site, through your CDN, and reads your sitemap from robots.txt when you add the site. The fetcher sends tollbot/1.0 from two ${a(T.staticIps, 'published addresses')}. Every firewall, rate limit and bot shield in front of your origin has to let it in.` },
+ { title: 'Stream your logs.', text: `The Platform overview requires them: "Every site that onboards to TollBit is required to send us their HTTP logs" (${a(T.overview, 'Platform overview')}, §1.3). Your CDN streams them with a secret key, or TollBit reads a storage bucket, which "takes up to 24 hours" to verify (${a(T.architecture, 'Platform architecture')}). TollBit's Cloudflare guide says the paywall works without them (${a(T.cloudflare, 'Cloudflare')}).` },
+ { title: 'Add one forwarding rule at your CDN.', text: `It tests the User-Agent header, the name a bot sends, against 69 AI user agents (${a(T.cloudflare, 'Cloudflare')}). On Cloudflare and Fastly the rule proxies the request, so the bot gets TollBit's answer at your URL. On CloudFront it switches the origin. On Vercel, Imperva, Google Cloud and Azure it redirects, and TollBit advises "a 302 (temporary) redirect" (${a(T.other, 'Other')}). TollBit has guides for 12 platforms (${a(T.integrations, 'Integrations')}).` },
+ { title: 'Decide what each bot gets.', text: `Your copy of the list decides which bots are forwarded at all. On Fastly, DataDome and Netacea you can pick agents one by one in a console instead. In TollBit's dashboard you "define which agents are allowed in, which are blocked, and which are subject to your paywall" (${a(T.configure, 'Configure Agent Site')}). Open Access serves bots or paths you choose a cached copy for free (${a(T.cache, 'Content cache')}). By default, any other forwarded bot without a pass gets a short message with HTTP 402, "payment required", and no price in it (${a(T.defaults, 'Default settings')}).` },
+];
+
+const CAUTIONS = [
+ { title: 'A blocked fetcher breaks the paywall.', text: `"If our requests are blocked, challenged, or rate-limited, your Agent Site will not work correctly" (${a(T.verifying, 'Verifying a property')}).` },
+ { title: 'Readers\' requests change too.', text: `On the recommended CloudFront setup, "Your regular origin will receive requests with its own domain as the Host rather than your public domain" (${a(T.aws, 'AWS')}).` },
+ { title: 'No fallback, even when you leave.', text: `One CloudFront setup "cannot fall back to your origin" when the Agent Site is down, so visitors get an error (${a(T.aws, 'AWS')}). TollBit's docs have no offboarding page; see "Plan the exit", below.` },
+ { title: 'Bots that hide their name get through.', text: `"We primarily focus on the big, self-identified bots" (${a(T.botManagement, 'Bot management')}). A bot that isn't on the list, or sends a browser's name, reaches your origin.` },
+ { title: 'The bot list goes stale.', text: `TollBit updates it "every quarter", and "publishers would then manually update their edge configurations" on most platforms (${a(T.botManagement, 'Bot management')}).` },
+];
+
+const AHEAD = [
+ { title: 'A buyer integrates once.', text: `One account, one card and one search reach every TollBit site that shows that buyer a rate (${a(T.devOverview, 'Developer overview')}). A FORA agent signs with one identity everywhere, but opens an account at each Exchange it buys from (${a('/getting-started/what-is-fora/', 'What is FORA?')}).` },
+ { title: 'It runs on more of your platforms.', text: `TollBit has setup guides for 12 platforms: CDNs, clouds, hosts and bot-security vendors. On Arc XP, WordPress VIP, DataDome and Netacea, forwarding bots is a setting in that vendor's console (${a(T.integrations, 'Integrations')}). FORA's publisher console sets up two networks, Cloudflare and AWS CloudFront (${a('/getting-started/publisher-onboarding/', 'Publisher onboarding')}).` },
+ { title: 'You see the bots before any sale.', text: `TollBit's dashboard shows scrapes, visits and referrals by agent, from your logs (${a(T.analytics, 'Enable analytics')}). FORA Server records transactions, not bot traffic or searches. It's open source, so an operator can add logs of Broker queries and Exchange offers to see what agents look for.` },
+ { title: 'It prepares content for buyers.', text: `TollBit converts pages to Markdown, holds back wire copy and embargoed articles, and tells buyers about new pages by webhook (${a(T.overview, 'Platform overview')}, §3). FORA leaves delivery to your CDN by design, and has no change feed to agents.` },
+ { title: 'Its license carries a warranty and an indemnity.', text: `FORA cannot state either as a term. Its obligations are duties of the agent, so a warranty fits only as prose in a license document a term points to (${a('/protocol/licensing-terms/', 'Licensing terms')}).` },
+];
+
+const COMPARE = compareRows([
+ [`Nowhere on your site. A forwarded bot gets a 402 by default; your rates live in TollBit's dashboard ${cite(3, 21)}.`, `Your fora.json names the Exchanges that sell your content, and you set terms with each ${cite(27)}.`],
+ [`TollBit's search, which covers sites across the web and flags its own network ${cite(11)}. How the index is built isn't documented.`, `From your file and your Exchange, or through a Broker ${cite(27, 31)}. No index spans the network.`],
+ [`Bots: by user agent, from a list of 69 ${cite(8)}. Buyers: a TollBit account, bound into each pass, which only TollBit can check ${cite(1, 2, 4)}.`, `A signature on each request, checked against the buyer's published key ${cite(32)}. The edge also matches the names bots send ${cite(28)}.`],
+ [`A pass for one URL, five minutes, one use, checked by TollBit alone ${cite(2, 4)}.`, `A signed link that expires and can be bound to the agent's key. Any edge with the Exchange's public key can check it ${cite(27, 33)}.`],
+ [`You: per 1,000 pages, per license type, by bot, page, keyword, time and directory. US dollars ${cite(2, 3)}.`, `You: free, per unit (pages, text tokens, seconds or your own) or flat, limited by use, country and buyer type, with caps. A currency per price ${cite(27, 30)}.`],
+ [`Through Stripe, monthly. You keep your full rate; the buyer pays an unpublished fee on top ${cite(1, 3)}.`, FORA_CELLS.money],
+ [`Daily bot analytics, and each sale: agent, page, price ${cite(3, 4)}.`, `Each sale, in the publisher console your Exchange operator runs ${cite(26, 27)}. No protocol call returns your sales.`],
+ [`No dispute call, and no charge without a delivered page ${cite(1)}. Then New York courts, or arbitration with TollBit ${cite(5, 6)}.`, FORA_CELLS.disputes],
+]);
+
+const BOTH_STEPS = [
+ { title: 'FORA\'s buyers skip TollBit\'s rule.', text: `So do requests for your FORA file. TollBit's rule looks for the link's signature and expiry (${a('/components/edge-function/signed-url-verification/', 'Signed URL verification')}). FORA's edge function then checks the link and serves the page from your origin.` },
+ { title: 'TollBit\'s own fetcher goes to your origin.', text: `FORA's edge function refuses user agents with a word ending in "bot" on the paths you license, and tollbot/1.0 is one. Add it to the edge function's allow list, beside your search crawlers (FORA Server's ${a(T.edgeConfig, 'edge configuration')}).` },
+ { title: 'Other AI bots get FORA\'s refusal or TollBit\'s paywall,', text: 'path by path.' },
+ { title: 'Readers go to your origin.', text: '' },
+];
+
+const TODO = [
+ { title: 'Ask what a buyer pays.', text: 'You keep your full rate, and TollBit\'s fee on top isn\'t published. Ask for the total a buyer pays for a page at your rate. FORA itself is free, open-source software.' },
+ { title: 'Read the license you accept.', text: `Activating rates accepts TollBit's standard license, warranty included (${a(T.settingRates, 'Setting rates')}). Send the points below to legal.` },
+ { title: 'Count what you hand over.', text: 'A subdomain under your name, and your HTTP logs, with your readers\' requests in them. On most platforms your engineers also update the bot list by hand each quarter.' },
+ { title: 'Check your rule order.', text: 'Let TollBit\'s fetcher through, and keep its rule off FORA links.' },
+ { title: 'Plan the exit.', text: 'Remove the forwarding rule before the DNS records, or setups without a fallback show visitors errors.' },
+ { title: 'Ask about Web Bot Auth.', text: `TollBit logs the signature headers that some AI agents send with each request (${a(T.cloudflare, 'Cloudflare')}). Ask whether it verifies them.` },
+ { title: 'Sell to agents as well.', text: `Name your FORA Exchanges in fora.json, with terms TollBit's rate card has no field for: a price by country or buyer type, or per text token (${a(T.rates, 'Rates')}).` },
+];
+
+const LEGAL = [
+ { title: 'TollBit isn\'t a party to the license.', text: `The license is a PDF contract between you and each buyer: "TollBit is not a party" (${a(T.terms, 'Publisher Terms')}).` },
+ { title: 'You give a warranty and an indemnity.', text: `You warrant that you have "all rights necessary to provide the Site Content". You indemnify the buyer against third-party claims that the content infringes a US copyright or trade secret. The license bars training, and passing content on without "express written consent". New York law (${a(T.license, 'standard license')}, §1.3, §6.2, §8.1, §9.4).` },
+ { title: 'Disputes.', text: `TollBit has no dispute call. Disputes with a buyer go to the courts of New York; disputes with TollBit go to arbitration (${a(T.terms, 'Publisher Terms')}).` },
+ { title: 'Your logs carry readers\' data.', text: `They cover "both bot and human requests". For sites that joined from Q4 2025 they include IP addresses, after TollBit revised its Data Processing Agreement (${a(T.overview, 'Platform overview')}, §1.3).` },
+ { title: 'Exit waits for active licenses.', text: `You can end the Publisher Terms only "if you are not a party to an active Content License Agreement" (${a(T.terms, 'Publisher Terms')}).` },
+ { title: 'Versions.', text: 'Publisher Terms updated October 29, 2025; standard license "Last Updated: September 23, 2024". TollBit can change both.' },
+];
+
+// The SourceList, numbered in order of first citation: an entry's position,
+// counted across the groups, is the N in cite(N).
+const SOURCES = [
+ {
+ name: 'TollBit',
+ items: [
+ { title: 'Platform overview', publisher: 'TollBit', href: T.overview, checked: '2026-11-14' },
+ { title: 'Tokens', publisher: 'TollBit docs', href: T.tokens, checked: '2026-11-14' },
+ { title: 'Setting rates', publisher: 'TollBit docs', href: T.settingRates, checked: '2026-11-15' },
+ { title: 'Security', publisher: 'TollBit docs', href: T.security, checked: '2026-11-15' },
+ { title: 'Publisher Terms of Service', publisher: 'TollBit', href: T.terms, checked: '2026-11-14' },
+ { title: 'Content License Agreement (standard license)', publisher: 'TollBit', href: T.license, checked: '2026-11-14' },
+ { title: 'Verifying a property', publisher: 'TollBit docs', href: T.verifying, checked: '2026-11-15' },
+ { title: 'Cloudflare', publisher: 'TollBit docs', href: T.cloudflare, checked: '2026-11-15' },
+ { title: 'Integrations', publisher: 'TollBit docs', href: T.integrations, checked: '2026-11-15' },
+ { title: 'Bot management', publisher: 'TollBit docs', href: T.botManagement, checked: '2026-11-15' },
+ { title: 'Licensed search', publisher: 'TollBit docs', href: T.licensedSearch, checked: '2026-11-14' },
+ { title: 'Rates', publisher: 'TollBit docs', href: T.rates, checked: '2026-11-14' },
+ { title: 'TollBit and Akamai partnership', publisher: 'TollBit blog', href: T.akamai, checked: '2026-11-14' },
+ { title: 'Announcing our Series A', publisher: 'TollBit blog', href: T.seriesA, checked: '2026-11-14' },
+ { title: 'TollBit homepage', publisher: 'TollBit', href: T.home, checked: '2026-11-14' },
+ { title: 'Static IPs', publisher: 'TollBit', href: T.staticIps, checked: '2026-11-14' },
+ { title: 'Platform architecture', publisher: 'TollBit docs', href: T.architecture, checked: '2026-11-15' },
+ { title: 'Other platforms', publisher: 'TollBit docs', href: T.other, checked: '2026-11-15' },
+ { title: 'Configure Agent Site', publisher: 'TollBit docs', href: T.configure, checked: '2026-11-15' },
+ { title: 'Content cache', publisher: 'TollBit docs', href: T.cache, checked: '2026-11-15' },
+ { title: 'Default settings', publisher: 'TollBit docs', href: T.defaults, checked: '2026-11-15' },
+ { title: 'AWS (S3, CloudFront, ALB)', publisher: 'TollBit docs', href: T.aws, checked: '2026-11-15' },
+ { title: 'Developer overview', publisher: 'TollBit docs', href: T.devOverview, checked: '2026-11-14' },
+ { title: 'Enable analytics', publisher: 'TollBit docs', href: T.analytics, checked: '2026-11-15' },
+ { title: 'Content (API)', publisher: 'TollBit docs', href: T.content, checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'FORA',
+ items: [
+ { title: 'Publisher onboarding', publisher: 'FORA documentation', href: '/getting-started/publisher-onboarding/', checked: '2026-11-15' },
+ { title: 'What is FORA?', publisher: 'FORA documentation', href: '/getting-started/what-is-fora/', checked: '2026-11-15' },
+ { title: 'Bot detection', publisher: 'FORA documentation', href: '/components/edge-function/bot-detection/', checked: '2026-11-15' },
+ { title: 'Edge function', publisher: 'FORA documentation', href: '/components/edge-function/overview/#forajson-and-rsltxt-pass-through', checked: '2026-11-15' },
+ { title: 'Licensing terms', publisher: 'FORA documentation', href: '/protocol/licensing-terms/', checked: '2026-11-15' },
+ { title: 'Broker', publisher: 'FORA documentation', href: '/components/broker/overview/', checked: '2026-11-15' },
+ { title: 'Authentication', publisher: 'FORA documentation', href: '/protocol/authentication/', checked: '2026-11-15' },
+ { title: 'Composition', publisher: 'FORA documentation', href: '/components/edge-function/composition/', checked: '2026-11-15' },
+ { title: 'Signed URL verification', publisher: 'FORA documentation', href: '/components/edge-function/signed-url-verification/', checked: '2026-11-15' },
+ { title: 'FORA Server edge configuration', publisher: 'FORA-Protocol on GitHub', href: T.edgeConfig, checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Press, filings and others',
+ items: [
+ { title: 'TNL Mediagene announces early success in AI content licensing', publisher: 'PR Newswire', href: T.tnl, checked: '2026-11-14' },
+ { title: 'The Washington Post\'s Arc XP adds TollBit', publisher: 'Digiday', href: T.digiday, checked: '2026-11-14' },
+ ],
+ },
+];
+---
+
+
+ Where TollBit is ahead
+ Run both
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
TollBit's rule forwards only the AI bots on its list. Googlebot and Bingbot aren't on it, so they reach your site as today.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
On one site
+
+
+
+
+
+
Behind a FORA Broker
+
+
Doesn't fit
+
Settling a FORA sale through TollBit. TollBit's API bills the calling account at TollBit's own rates, even for usage the buyer reports itself. It has no call that bills a price another Exchange set (${a(T.content, 'Content')}).`} />
+
+
+
+
+
+
+
+
+
+
+
+
+
Sources
+
Sources are checked on the date of publication.
+
+
+
+
+
+
+ What is FORA?
+ Run your own marketplace
+ All content marketplaces
+ Licensing terms
+
+
+
+
diff --git a/website/src/pages/resources/index.astro b/website/src/pages/resources/index.astro
index 956c1245..67c8e04b 100644
--- a/website/src/pages/resources/index.astro
+++ b/website/src/pages/resources/index.astro
@@ -1,11 +1,17 @@
---
-// /resources/: one card per topic, from src/data/resources.ts.
+// /resources/: one card per topic, from src/data/resources.ts. Topics without
+// a section first, then one band per section, with its heading and intro.
import LandingLayout from '../../layouts/LandingLayout.astro';
import AudienceHero from '../../components/landing/AudienceHero.astro';
import Band from '../../components/landing/Band.astro';
import ResourceCards from '../../components/landing/ResourceCards.astro';
import ClosingCta from '../../components/landing/ClosingCta.astro';
-import { RESOURCES } from '../../data/resources';
+import { RESOURCES, SECTIONS } from '../../data/resources';
+
+const loose = RESOURCES.filter((r) => !r.section);
+const sections = SECTIONS
+ .map((s) => ({ ...s, items: RESOURCES.filter((r) => r.section === s.name) }))
+ .filter((s) => s.items.length > 0);
---
-
+
+ {sections.map((s) => (
+
+
+
+ ))}
+
`${text}`;
+
+// The standards' own pages, linked from the table's first column.
+const S = {
+ robots: 'https://www.rfc-editor.org/rfc/rfc9309',
+ aipref: 'https://datatracker.ietf.org/wg/aipref/about/',
+ tdmrep: 'https://www.w3.org/community/reports/tdmrep/CG-FINAL-tdmrep-20240510/',
+ rsl: 'https://rslstandard.org/rsl',
+ comp: 'https://github.com/IABTechLab/CoMP',
+ llms: 'https://llmstxt.org/',
+ wba: 'https://datatracker.ietf.org/doc/draft-ietf-webbotauth-httpsig-protocol/',
+ mcp: 'https://modelcontextprotocol.io/specification/2026-07-28',
+ c2pa: 'https://c2pa.org/',
+ x402: 'https://www.x402.org/',
+ mpp: 'https://mpp.dev/',
+ spur: 'https://github.com/SPUR-Coalition/telemetry',
+ socw: 'https://m4ow.uk/socw/',
+};
+
+const STANDARDS = [
+ { name: 'robots.txt', href: S.robots, job: 'Declare terms: allow or block, per crawler', status: `IETF standard, RFC 9309 (2022). Its rules "are not a form of access authorization" ${cite(1)}`, users: `On sites using TollBit, 30% of AI bot scrapes in Q4 2025 ignored it (TollBit's own data) ${cite(2)}` },
+ { name: 'AIPREF', href: S.aipref, job: 'Declare terms: a preference about AI training and AI use', status: `IETF working-group draft -08 (September 2026), without group consensus yet; the robots.txt training token is an individual draft ${cite(4, 5)}`, users: `Anthropic has recognized the training token since October 30, 2026; Microsoft from February 1, 2027; Google intends to in early 2027; OpenAI "will honor" it ${cite(6, 7, 8, 9)}` },
+ { name: 'TDMRep', href: S.tdmrep, job: 'Declare terms: reserve text-and-data-mining rights', status: `W3C Community Group final report (2024), "not a W3C Standard" ${cite(11)}`, users: `More than 50 adopters on the group's own list, among them Hachette Livre, Elsevier, Springer Nature, Wiley and Radio France ${cite(12)}` },
+ { name: 'RSL', href: S.rsl, job: 'Declare terms: state license terms and prices; it also specifies a license server (OLP)', status: `RSL 1.0, a Recommendation of RSL's Technical Steering Committee, December 2025 ${cite(14, 15)}`, users: `More than 3,000 supporters, says ILC; one AI company licensed through ILC, Yahoo Scout ${cite(16)}` },
+ { name: 'IAB Tech Lab CoMP', href: S.comp, job: 'Declare terms: describe content, its price and where its license is, for a marketplace API', status: `IAB Tech Lab, final V1, April 2026 ${cite(21)}`, users: `FORA carries CoMP: its fields travel whole on each FORA offer ${cite(22)}` },
+ { name: 'llms.txt', href: S.llms, job: 'Help discovery: a curated index for AI readers', status: `A proposal by Jeremy Howard, v2 August 2026; no standards body ${cite(31)}`, users: `"thousands of sites publish an llms.txt file", by its author's count ${cite(31)}` },
+ { name: 'Web Bot Auth', href: S.wba, job: 'Identify the caller: a signature on each agent request', status: `IETF working-group draft, October 2026 ${cite(32)}`, users: `ChatGPT Work's cloud browser signs with it ${cite(33)}. Cloudflare verifies, and FORA agents sign, under two different earlier drafts ${cite(23, 34)}` },
+ { name: 'MCP', href: S.mcp, job: 'Connect: how an agent calls a tool or a seller', status: `The Agentic AI Foundation, under the Linux Foundation; revision of July 28, 2026 ${cite(35, 36)}`, users: `Cashmere, AskNews and Newstex sell content through MCP servers; FORA's Identity Service is one ${cite(22, 37, 38, 39)}` },
+ { name: 'C2PA', href: S.c2pa, job: 'Prove origin: who made this, and has it changed', status: `C2PA 2.5 (October 2026), a Linux Foundation project; ISO 22144 at committee draft ${cite(41, 42, 43)}`, users: `Steering-committee members include Adobe, BBC, Google, Meta, Microsoft and OpenAI ${cite(44)}` },
+ { name: 'FORA', fora: true, job: 'Declare terms, transact and report: a signed offer per term, the buyer\'s signed acceptance, a signed link, and the buyer\'s usage report', status: `The open protocol this site describes. Apache-2.0, release v1.2.1, with FORA Server, an open-source implementation ${cite(22, 24)}`, users: 'Any company can run its own marketplace on it. See About FORA' },
+ { name: 'x402', href: S.x402, job: 'Settle, which moves money: payment over HTTP 402', status: `x402 Foundation, under the Linux Foundation; 40 members by July 2026, Visa, Mastercard and American Express among them ${cite(48)}`, users: `In AWS WAF on CloudFront since June 2026; in Cloudflare's Monetization Gateway, a closed beta since September 30, 2026 ${cite(49, 50)}` },
+ { name: 'MPP', href: S.mpp, job: 'Settle, which moves money: payment over HTTP 402, per charge, session or subscription', status: `Co-authored by Tempo and Stripe; its core scheme is an individual Internet-Draft ${cite(52, 53)}`, users: `Cloudflare's Agents SDK names it as one of its two payment protocols; AWS called MPP support "coming soon" in June 2026 ${cite(49, 54)}` },
+ { name: 'SPUR Content Telemetry', href: S.spur, job: 'Report: what an AI system did with content after the fetch', status: `SPUR Coalition, 1.0, September 2026 ${cite(55)}`, users: `AskNews says it weights its publisher pool by its events ${cite(56)}` },
+];
+
+const KEY_FACTS = [
+ { value: '30%', text: `of AI bot scrapes ignored robots.txt on sites using TollBit, in Q4 2025, by TollBit's own data ${cite(2)}` },
+ { value: '3,000+', text: `RSL supporters, by ILC's count ${cite(16)}` },
+ { value: '4', text: `AI companies committed to honor AIPREF's training refusal ${cite(6, 7, 8, 9)}` },
+ { value: '1', text: `AI company paying under an RSL-based license: Yahoo Scout ${cite(16)}` },
+];
+
+const STANDING = [
+ { title: 'Blocking is a request, and a growing share of AI bot scrapes ignore it.', text: `On sites using TollBit, the share that bypassed robots.txt rose from 3.3% in Q4 2024 to 12.9% in Q1 2025, then to 30% ${cite(2, 10)}. AIPREF says so of itself: a receiver "has a choice whether to follow those preferences" ${cite(4)}.` },
+ { title: 'Declaring has thousands of supporters, four committed AI companies and one buyer.', text: `The four commitments promise to honor a refusal, and carry no payment. Yahoo Scout pays under ILC's blanket license, announced on November 12, 2026 ${cite(16)}. No AI company is known to pay under a publisher's own RSL terms.` },
+ { title: 'Transacting and settling are available in a few places.', text: `FORA transacts on every Exchange that runs it. x402 settles in AWS WAF and Cloudflare's closed beta, MPP in Cloudflare's Agents SDK. Supertab sells an RSL license server, and no AI company is named as buying through it ${cite(25, 49, 50, 54, 57)}.` },
+];
+
+const TODO = [
+ { title: 'Write down what training, grounding and display are each worth to you.', text: 'Grounding means an AI answer uses your page as its source. None of these standards can tell you that.' },
+ { title: 'Generate your rules.', text: `The ${a('/resources/ai-crawlers/#generator', 'AI rules generator')} writes robots.txt groups for AI crawlers, AIPREF's training refusal, the line that attaches the search-only contract, and TDMRep's header and tag.` },
+ { title: 'Adopt the search-only contract, with your legal team.', text: `Set your fee in your site's terms and point robots.txt at the contract: ${a('/getting-started/search-only-contract/', 'the steps')}.` },
+ { title: 'Publish RSL if you have license terms to state.', text: `Behind FORA's edge function, keep it at /rsl.txt, which the edge passes through to every client ${cite(27)}.` },
+ { title: 'See what an agent would be offered.', text: `Paste one of your pages on ${a('/publishers/', 'For Publishers')}. No account needed.` },
+];
+
+const LEGAL = [
+ { title: 'Formed by conduct.', text: `Under the contract, "agreement is formed through conduct, simply by accessing" the site ${cite(63)}.` },
+ { title: 'Law.', text: `Under the law of England and Wales, the fee is "a contractual debt which becomes immediately due" ${cite(63)}. Whether it holds where you publish is a question for your lawyer.` },
+];
+
+const DIFFER = [
+ { title: 'A price and a credit together.', text: `In RSL, a license asks for money or for credit, not both. Its schema (RSL 1.0, Appendix A) allows one payment element with one type, and attribution is one of the types ${cite(14)}. CoMP keeps citation in a field of its own, beside the price ${cite(20)}. A FORA term can charge per access and require credit ${cite(28)}.` },
+ { title: 'Caps.', text: `FORA states a cap as a metric, a limit and a window: ten seats a month, say. The buyer agrees to it; the Exchange doesn't meter it ${cite(28)}. RSL's schema has no cap element (Appendix A) ${cite(14)}. CoMP has a yes-or-no field that says a crawl limit exists ${cite(20)}.` },
+ { title: 'A menu for one asset.', text: `A FORA resource carries up to 32 terms, and each one a buyer may see is its own signed offer ${cite(28)}. RSL allows several licenses for the same content, but tells processors to "evaluate all applicable licensing terms together" ${cite(14)}. A CoMP response carries one package with one price ${cite(20)}.` },
+ { title: 'Saying what you want first.', text: `CoMP lets an AI system state its intended use ${cite(20)}. FORA lets a buyer state the uses, places and kind of user it accepts, and an Exchange may return only the offers that fit ${cite(28)}.` },
+];
+
+const EXPLORER = [
+ { title: 'Price and credit.', text: 'Pick "AI answers, pay per crawl": RSL loses the credit, CoMP keeps it.' },
+ { title: 'Caps and reports.', text: 'Pick "Annual AI access deal": the cap is in neither file.' },
+ { title: 'A shared license.', text: 'Pick "Collective blanket license": neither RSL nor CoMP can pin the document\'s version.' },
+];
+
+const ENGINEERS = [
+ { title: 'Pairs that compose.', text: `RSL adds a License: line to robots.txt; AIPREF's training token is a robots.txt group ${cite(5, 14)}. x402's "exact" flows "map directly onto MPP's charge intent", and MPP can charge for MCP tool calls ${cite(52)}. MCP defines no price or license: each seller adds its own ${cite(35)}.` },
+];
+
+const FAQ = [
+ { q: 'Do I have to choose between RSL and FORA?', a: 'No. FORA doesn\'t replace RSL or CoMP. RSL publishes your terms in a file anyone can read. FORA sells under the terms you set in the publisher console, one signed offer at a time.' },
+ { q: 'Is TDMRep enough to reserve AI training in the EU?', a: `TDMRep is a machine-readable form of the reservation the EU's copyright directive recognizes. In the TDMRep report's summary of Article 4, mining is allowed unless "expressly reserved by their rightsholders in an appropriate manner, such as machine-readable means" ${cite(11)}. Whether yours meets the test is a question for your lawyer.` },
+ { q: 'What isn\'t on the map?', a: `IPTC's photo metadata has a data-mining property inside the image file ${cite(60)}. ODRL is the policy language TDMRep uses ${cite(11, 61)}. The European Commission's consultation lists Liccium's TDM·AI among opt-out protocols. The Commission hasn't published its list of generally agreed ones ${cite(62)}.` },
+];
+
+// The SourceList, numbered in order of first citation: an entry's position,
+// counted across the groups, is the N in cite(N).
+const SOURCES = [
+ {
+ name: 'robots.txt and AIPREF',
+ items: [
+ { title: 'RFC 9309: Robots Exclusion Protocol', publisher: 'IETF', href: S.robots, checked: '2026-11-14' },
+ { title: 'State of the Bots, Q3–Q4 2025', publisher: 'TollBit', href: 'https://tollbit.com/state-of-the-bots/q3-q4-2025/', checked: '2026-11-14' },
+ { title: 'AI Preferences (aipref) working group', publisher: 'IETF', href: S.aipref, checked: '2026-11-15' },
+ { title: 'A Vocabulary For Expressing AI Usage Preferences, draft -08', publisher: 'IETF AIPREF working group', href: 'https://datatracker.ietf.org/doc/draft-ietf-aipref-vocab/', checked: '2026-11-15' },
+ { title: 'Machine-Readable Training Preference Signal', publisher: 'IETF, individual Internet-Draft', href: 'https://datatracker.ietf.org/doc/draft-kelley-aipref-training-preference-signal/', checked: '2026-11-15' },
+ { title: 'robots.txt controls', publisher: 'Anthropic', where: 'Anthropic help center, article 8896518 (no link)', checked: '2026-11-15' },
+ { title: 'Your content, your choice: clearer preferences for training and grounding', publisher: 'Microsoft, Bing blog', href: 'https://blogs.bing.com/search/2026/10/your-content-your-choice-clearer-preferences-for-training-and-grounding/', checked: '2026-11-15' },
+ { title: 'Google\'s common crawlers', publisher: 'Google for Developers', href: 'https://developers.google.com/crawling/docs/crawlers-fetchers/google-common-crawlers', checked: '2026-11-15' },
+ { title: 'Helping websites express AI training preferences', publisher: 'OpenAI', href: 'https://openai.com/index/helping-websites-express-ai-training-preferences/', checked: '2026-11-14' },
+ { title: 'State of the Bots, Q1 2025', publisher: 'TollBit', href: 'https://tollbit.com/state-of-the-bots/q1-2025/', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'TDMRep',
+ items: [
+ { title: 'TDM Reservation Protocol (TDMRep), Final Community Group Report', publisher: 'W3C TDMRep Community Group', href: S.tdmrep, checked: '2026-11-14' },
+ { title: 'TDMRep adopters', publisher: 'W3C TDMRep Community Group, on GitHub', href: 'https://github.com/w3c/tdm-reservation-protocol/blob/main/docs/adopters.md', checked: '2026-11-14' },
+ { title: 'Text and Data Mining Reservation Protocol Community Group', publisher: 'W3C', href: 'https://www.w3.org/community/tdmrep/', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'RSL and ILC',
+ items: [
+ { title: 'Really Simple Licensing (RSL) 1.0 specification', publisher: 'RSL Technical Steering Committee', href: S.rsl, checked: '2026-11-14' },
+ { title: 'RSL AI Licensing 1.0 now an official industry standard', publisher: 'RSL, press release', href: 'https://rslstandard.org/press/rsl-1-specification-2025', checked: '2026-11-14' },
+ { title: 'The Internet Licensing Collective launches with Yahoo Scout as first commercial partner', publisher: 'Internet Licensing Collective, press release', href: 'https://internetcollective.org/press/ilc-launch-2026', checked: '2026-11-14' },
+ { title: 'RSL quickstart', publisher: 'RSL', href: 'https://rslstandard.org/', checked: '2026-11-14' },
+ { title: 'The ILC Platform', publisher: 'Internet Licensing Collective', href: 'https://internetcollective.org/platform', checked: '2026-11-13' },
+ { title: 'ILC\'s RSL file', publisher: 'Internet Licensing Collective', href: 'https://internetcollective.org/rsl.xml', checked: '2026-11-13' },
+ ],
+ },
+ {
+ name: 'CoMP',
+ items: [
+ { title: 'CoMP 1.0: Content Metadata Marketplace Supply Specification', publisher: 'IAB Tech Lab, on GitHub', href: S.comp, checked: '2026-11-14' },
+ { title: 'CoMP: Content Monetization Protocols initiative', publisher: 'IAB Tech Lab', href: 'https://iabtechlab.com/standards/comp-content-monetization-protocols-initiative/', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'FORA',
+ items: [
+ { title: 'What Is FORA?', publisher: 'FORA documentation', href: '/getting-started/what-is-fora/', checked: '2026-11-15' },
+ { title: 'Authentication', publisher: 'FORA documentation', href: '/protocol/authentication/', checked: '2026-11-15' },
+ { title: 'FORA protocol repository', publisher: 'FORA-Protocol on GitHub', href: 'https://github.com/FORA-Protocol/protocol', checked: '2026-11-15' },
+ { title: 'Broker', publisher: 'FORA documentation', href: '/components/broker/overview/', checked: '2026-11-15' },
+ { title: 'Search-Only Contract', publisher: 'FORA documentation', href: '/getting-started/search-only-contract/', checked: '2026-11-15' },
+ { title: 'FORA Server', publisher: 'FORA-Protocol on GitHub', href: 'https://github.com/FORA-Protocol/reference-implementation', checked: '2026-11-15' },
+ { title: 'Licensing Terms', publisher: 'FORA documentation', href: '/protocol/licensing-terms/', checked: '2026-11-15' },
+ { title: 'The licensing core explorer', publisher: 'FORA', href: '/protocol/#licensing-core', checked: '2026-11-15' },
+ { title: 'Publisher Onboarding', publisher: 'FORA documentation', href: '/getting-started/publisher-onboarding/', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'llms.txt',
+ items: [
+ { title: 'The /llms.txt file', publisher: 'Jeremy Howard, llmstxt.org', href: S.llms, checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'Web Bot Auth',
+ items: [
+ { title: 'HTTP Message Signatures for automated traffic (draft-ietf-webbotauth-httpsig-protocol)', publisher: 'IETF webbotauth working group', href: S.wba, checked: '2026-11-14' },
+ { title: 'ChatGPT Work\'s cloud browser allowlisting', publisher: 'OpenAI Help Center', href: 'https://help.openai.com/en/articles/11845367', checked: '2026-11-14' },
+ { title: 'Web Bot Auth', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'MCP',
+ items: [
+ { title: 'Model Context Protocol specification, revision 2026-07-28', publisher: 'Model Context Protocol', href: S.mcp, checked: '2026-11-15' },
+ { title: 'MCP joins the Agentic AI Foundation', publisher: 'Model Context Protocol blog', href: 'https://blog.modelcontextprotocol.io/posts/2025-12-09-mcp-joins-agentic-ai-foundation/', checked: '2026-11-15' },
+ { title: 'MCP server documentation', publisher: 'Cashmere', href: 'https://cashmere.io/docs/mcp', checked: '2026-11-14' },
+ { title: 'MCP server', publisher: 'AskNews docs', href: 'https://docs.asknews.app/en/mcp', checked: '2026-11-15' },
+ { title: 'Newstex MCP Server', publisher: 'Newstex', href: 'https://mcp.newstex.com/', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'C2PA',
+ items: [
+ { title: 'C2PA: Coalition for Content Provenance and Authenticity', publisher: 'C2PA', href: S.c2pa, checked: '2026-11-14' },
+ { title: 'C2PA Technical Specification 2.5', publisher: 'C2PA', href: 'https://spec.c2pa.org/specifications/specifications/2.5/specs/C2PA_Specification.html', checked: '2026-11-14' },
+ { title: 'About C2PA', publisher: 'C2PA', href: 'https://c2pa.org/about/', checked: '2026-11-14' },
+ { title: 'ISO/CD 22144, Authenticity of information: Content credentials', publisher: 'ISO', href: 'https://www.iso.org/standard/90726.html', checked: '2026-11-14' },
+ { title: 'Membership', publisher: 'C2PA', href: 'https://c2pa.org/membership/', checked: '2026-11-14' },
+ { title: 'C2PA founding press release', publisher: 'C2PA', href: 'https://c2pa.org/c2pa-founding-press-release/', checked: '2026-11-14' },
+ { title: 'Training and Data Mining Assertion 1.1', publisher: 'Creator Assertions Working Group', href: 'https://cawg.io/training-and-data-mining/1.1/', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'Payments: x402 and MPP',
+ items: [
+ { title: 'x402', publisher: 'x402 Foundation', href: S.x402, checked: '2026-11-14' },
+ { title: 'Linux Foundation announces operational launch of x402 Foundation', publisher: 'x402 Foundation', href: 'https://x402.org/linux-foundation-announces-operational-launch-of-x402-foundation-to-standardize-internet-native-payments-for-ai-agents-and-applications/', checked: '2026-11-14' },
+ { title: 'AWS WAF announces AI traffic monetization', publisher: 'AWS, What\'s New', href: 'https://aws.amazon.com/about-aws/whats-new/2026/06/aws-waf-ai-traffic-monetization/', checked: '2026-11-14' },
+ { title: 'Monetization Gateway beta: charge AI agents for consumption with HTTP 402', publisher: 'Cloudflare blog', href: 'https://blog.cloudflare.com/monetization-gateway-beta/', checked: '2026-11-14' },
+ { title: 'Machine Payments Protocol', publisher: 'MPP', href: S.mpp, checked: '2026-11-15' },
+ { title: 'Frequently asked questions', publisher: 'MPP', href: 'https://mpp.dev/faq', checked: '2026-11-15' },
+ { title: 'The "Payment" HTTP Authentication Scheme', publisher: 'IETF, individual Internet-Draft', href: 'https://datatracker.ietf.org/doc/draft-httpauth-payment/', checked: '2026-11-15' },
+ { title: 'Agentic Payments, Agents SDK', publisher: 'Cloudflare docs', href: 'https://developers.cloudflare.com/agents/tools/payments/', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'SPUR Content Telemetry',
+ items: [
+ { title: 'SPUR Content Telemetry 1.0', publisher: 'SPUR Coalition, on GitHub', href: S.spur, checked: '2026-11-14' },
+ { title: 'From Retrieval to Citation: How AskNews Makes Publisher Usage Traceable', publisher: 'AskNews blog', href: 'https://asknews.app/en/blog-articles/spur-telemetry-publisher-usage-and-revenue', checked: '2026-11-14' },
+ ],
+ },
+ {
+ name: 'Others mentioned',
+ items: [
+ { title: 'Supertab Connect', publisher: 'Supertab', href: 'https://www.supertab.co/', checked: '2026-11-13' },
+ { title: 'Daily Kos\'s RSL file', publisher: 'Daily Kos', href: 'https://www.dailykos.com/license.xml', checked: '2026-11-13' },
+ { title: 'Tokens', publisher: 'TollBit docs', href: 'https://docs.tollbit.com/docs/tokens', checked: '2026-11-14' },
+ { title: 'IPTC Photo Metadata Standard', publisher: 'IPTC', href: 'https://iptc.org/std/photometadata/specification/IPTC-PhotoMetadata', checked: '2026-11-15' },
+ { title: 'ODRL Information Model 2.2', publisher: 'W3C', href: 'https://www.w3.org/TR/odrl-model/', checked: '2026-11-15' },
+ { title: 'Stakeholder consultation on AI and copyright compliance', publisher: 'European Commission', href: 'https://digital-strategy.ec.europa.eu/en/faqs/stakeholder-consultation-ai-and-copyright-compliance', checked: '2026-11-15' },
+ ],
+ },
+ {
+ name: 'Contracts: the search-only contract',
+ items: [
+ { title: 'Search Only Terms Contract', publisher: 'Movement for an Open Web', href: S.socw, checked: '2026-11-14' },
+ ],
+ },
+];
+---
+
+
+ See the thirteen
+ Generate your rules
+
+
+
+
+
A machine-readable declaration is a request. TDMRep reserves rights that bind in the EU, and no technical layer enforces them. A declaration earns money only when a license server, an Exchange or a contract that access triggers stands behind it.
+
+
Ask your vendors. Put this to your CDN, bot-management vendor or marketplace, verbatim: "Which of these standards do you read, which do you enforce at the door, and who is paying under any of them through you today?"
+
+
+
+
+
+
+
+
+
How to read it. Each row does a different job, so the rows don't rank against each other: a road sign isn't worse than a till for not taking payment. The last column mixes kinds of adoption, and each cell says which.
+
Transacting agrees one sale; settling moves the money. RSL declares terms, and its specification describes a license server. FORA spans the terms, the transaction and the report, and leaves the money to your operator.
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
license server included, carries the buyer's own signature on the terms it accepted.`} />
+
+
+
+
+
+
The rest of the page compares the specifications clause by clause. Everything above reads without it.
+
+
+
+
+
Three of the thirteen are license formats: RSL 1.0, CoMP and FORA's licensing core. All three say what you may do, who may buy, where, at what price, and what you owe afterward. TDMRep and AIPREF say only the first.
+
all to search ${cite(14)}. CoMP takes the same six names, four definitions word for word ${cite(20)}. FORA starts from RSL's use words and adds copyright uses such as reproduce and sync, 23 in all. It also accepts AIPREF's train-ai as a spelling of ai-train ${cite(28)}. RSL and FORA both scope a term by place and kind of user, and both can prohibit as well as permit ${cite(14, 28)}.`} />
+
+
+
+
+
<standard> element points at a legal text elsewhere, such as a Creative Commons license ${cite(14)}. FORA incorporates by reference too, with the document's digest beside its address, so the document can't be swapped after an offer is signed ${cite(28)}.`} />
+
+
See it clause by clause. The ${a('/protocol/#licensing-core', 'Licensing Core explorer')} shows 12 licenses, each as a FORA term, an RSL file and a CoMP package.`} />
+
+
+
+
+
+
+
+
+
<standard> takes an address with no digest (Appendix A) ${cite(14)}. TollBit's content tokens name a license type and a license ID, and TollBit signs them, not the buyer ${cite(59)}. CoMP records how a crawler plans to identify itself, which is not its assent ${cite(20)}.`} />
+
What ILC does with it. ILC, the collective that co-authored RSL, keeps the transaction out of the content path ${cite(16, 18)}. Its RSL file names no license server and points, through <standard>, at ILC's collective license ${cite(19)}. An AI company takes one blanket license, for a fee ILC doesn't publish, and reports its own use ${cite(18)}. Yahoo Scout, ILC's first licensee, pays under that contract ${cite(16)}. ${a('/resources/transparent-collective/', 'How ILC compares with other collectives →')}`} />
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Two pairs that can contradict each other. RSL states which uses are permitted; CoMP leaves uses to the license it links, so the two can disagree unseen ${cite(14, 20)}. If you publish both, check that they say the same thing. robots.txt governs fetching by crawler name and TDMRep reserves a use. The TDMRep group calls them complementary, but "this complementarity is not specified in the TDMRep document and remains subject to interpretation" ${cite(13)}.`} />
+
+
+
+
+
+
+
+
About FORA.`} />
+
+
+
+
+
+
+
+
+
+
Sources
+
+ All sources
+
+
+
Sources are checked on the date of publication.
+
+
+
+
+
+ Generate your rules
+ See your offers
+
+
+
+
+
+
diff --git a/website/src/pages/resources/transparent-collective.astro b/website/src/pages/resources/transparent-collective.astro
index a56e6e18..1a4ea864 100644
--- a/website/src/pages/resources/transparent-collective.astro
+++ b/website/src/pages/resources/transparent-collective.astro
@@ -628,7 +628,7 @@ const SOURCES = [
.scheme { padding: 1.75rem; border: 1px solid var(--lp-rule); border-radius: 12px; background: #fff; min-width: 0; }
.scheme.wide, .scheme:last-child:nth-child(even) { grid-column: 1 / -1; }
.scheme p { margin: 0; overflow-wrap: break-word; }
- .scheme .scheme-tag { font: 500 0.875rem var(--mono); letter-spacing: 0.06em; text-transform: uppercase; line-height: 1.4; color: var(--lp-link); margin-bottom: 0.625rem; }
+ .scheme .scheme-tag { font: 500 0.875rem var(--mono); letter-spacing: 0.06em; text-transform: uppercase; line-height: 1.4; color: var(--lp-label); margin-bottom: 0.625rem; }
.scheme.wide p:not(.scheme-tag), .scheme:last-child:nth-child(even) p:not(.scheme-tag) { max-width: 46rem; }
.world { display: grid; grid-template-columns: repeat(2, minmax(0, 1fr)); gap: 1.25rem; align-items: start; margin-top: 2.5rem; }
.world-region { padding: 1.5rem 1.75rem; border: 1px solid var(--lp-rule); border-radius: 12px; background: #fff; min-width: 0; }
@@ -638,7 +638,7 @@ const SOURCES = [
.world-region li { display: grid; gap: 0.25rem; color: var(--lp-text); font-size: 1.0625rem; line-height: 1.6; overflow-wrap: break-word; }
.world-region b { color: var(--lp-strong); font-weight: 700; }
.world-region :global(a) { color: var(--lp-link); }
- .world-tag { font: 500 0.875rem var(--mono); letter-spacing: 0.06em; text-transform: uppercase; color: var(--lp-link); }
+ .world-tag { font: 500 0.875rem var(--mono); letter-spacing: 0.06em; text-transform: uppercase; color: var(--lp-label); }
/* The same look as PointList, for lists whose text carries links. */
.points { list-style: none; margin: 2rem 0 0; padding: 0; display: grid; gap: 1.25rem; max-width: 44rem; }
.points li { display: grid; grid-template-columns: auto 1fr; gap: 0.75rem; color: var(--lp-text); font-size: 1.125rem; line-height: 1.7; }
diff --git a/website/src/pages/resources/web-bot-auth.astro b/website/src/pages/resources/web-bot-auth.astro
new file mode 100644
index 00000000..c771b239
--- /dev/null
+++ b/website/src/pages/resources/web-bot-auth.astro
@@ -0,0 +1,528 @@
+---
+// Resource: Web Bot Auth, how an AI agent proves who it is (RES-2). Its own
+// order: how it works (figure and four steps), what it changes for a
+// publisher, which agents sign, key directories with rotation and revocation,
+// what to ask your CDN and your team, choosing which agents you sell to, where
+// permissions come from, the debate, how FORA uses it, the detail for
+// engineers, the questions, and About FORA with the sources. Other identity
+// schemes, OAuth and MCP among them, are on /resources/agent-identity/.
+//
+// Shared texts (About FORA, the definitions, the search callout, the tagline)
+// come from src/data/content-marketplaces.ts. Copy is written with [N]
+// citation markers and passed through h(), which turns each into cite(N), a
+// link to entry N of the numbered SourceList; SOURCES is numbered as the copy
+// numbers it, so an entry's position is its number. Text with links is HTML
+// and goes through set:html.
+import LandingLayout from '../../layouts/LandingLayout.astro';
+import AudienceHero from '../../components/landing/AudienceHero.astro';
+import Band from '../../components/landing/Band.astro';
+import Callout from '../../components/landing/Callout.astro';
+import Figure from '../../components/landing/Figure.astro';
+import KeyFacts from '../../components/landing/KeyFacts.astro';
+import PullQuote from '../../components/landing/PullQuote.astro';
+import Split from '../../components/landing/Split.astro';
+import PointList from '../../components/landing/PointList.astro';
+import StepList from '../../components/landing/StepList.astro';
+import ResourceTable from '../../components/landing/ResourceTable.astro';
+import FaqGrid from '../../components/landing/FaqGrid.astro';
+import LinkRow from '../../components/landing/LinkRow.astro';
+import SourceList from '../../components/landing/SourceList.astro';
+import ClosingCta from '../../components/landing/ClosingCta.astro';
+import { resourceUpdated } from '../../data/resources';
+import { ABOUT_FORA, DEFINITIONS, SEARCH_FORA, TAGLINE } from '../../data/content-marketplaces';
+import { cite, statusTag } from '../../lib/resource-markup';
+
+const HREF = '/resources/web-bot-auth/';
+const IMG = '/resources/web-bot-auth';
+const ID = '/resources/agent-identity/';
+const a = (href: string, text: string) => `${text}`;
+const c = (s: string) => `${s}`;
+/** Turns each [N] or [N, M] marker in the copy into a citation link. */
+const h = (s: string) => s.replace(/\[(\d+(?:, \d+)*)\]/g, (_, g: string) => cite(...g.split(', ').map(Number)));
+
+// The two sources a PullQuote links; the rest are in SOURCES.
+const S = {
+ draft01: 'https://www.ietf.org/archive/id/draft-ietf-webbotauth-httpsig-protocol-01.txt',
+ eff: 'https://www.eff.org/deeplinks/2026/06/free-and-open-web-under-attack-ietf',
+};
+
+const HOW = [
+ { text: h('The agent\'s operator creates a key pair. The private key stays with the agent. The public key goes into the key directory, a small file on the operator\'s website [1].') },
+ { text: h(`The agent signs each request and adds three headers. ${c('Signature-Agent')} names its website. ${c('Signature-Input')} lists what the signature covers, which key made it and when it expires. ${c('Signature')} carries the signature itself [1].`) },
+ { text: h(`To cover the body, the agent also sends ${c('Content-Digest')}, a hash of the body, and signs it with the rest. A changed body no longer matches its hash, and a changed hash breaks the signature [1, 2].`) },
+ { text: h(`Your site, or your CDN, fetches the key directory from the website ${c('Signature-Agent')} names, takes the key with the matching ID and checks the signature. If it verifies, a key that website publishes signed the request, and nothing the signature covers was changed on the way [1, 2].`) },
+];
+
+const KEY_DIRECTORY = {
+ tag: 'What is a key directory?',
+ text: h(`A small file on the agent's website, usually at ${c('/.well-known/http-message-signatures-directory')}, listing the public keys the agent signs with. Whoever controls the website controls the file, so the website stands behind the keys. The website's address is the agent's identity: the keys in the file can change, and sites still recognize the agent [1].`),
+};
+
+const TAKEAWAY = 'Web Bot Auth tells you which website an agent signs for, and only a holder of that website\'s keys can sign in its name. It\'s for agents that want to be known, and they\'re the ones you can sell to. A bot that wants to stay hidden simply doesn\'t sign.';
+
+const WHY = [
+ { title: 'You can set terms per agent.', text: h('Sell to it, slow it down, serve it differently or refuse it, by its website\'s address [1].') },
+ { title: 'You know who you\'re dealing with.', text: h('Anyone can type a user agent, and a cloud IP address is shared by everyone who rents from that cloud [1]. A signature names the website behind the agent.') },
+ { title: 'A leaked signature expires.', text: h('A signature works only at the site it was made for, and only until it expires: at most a day by the draft\'s advice, and often a minute [1, 12]. Cover the path and a hash of the body, and it fits one request. A leaked password or API key works wherever it\'s accepted until someone revokes it.') },
+ { title: 'A page you sell stays sold to one buyer.', text: h('On FORA, the link to a sold page works only for the agent that holds the buyer\'s key, so a copied link fetches nothing [7].') },
+ { title: 'No passwords or API keys to issue.', text: h('An agent can open an account on its first signed request, with no sign-up form. You keep only public keys, and a public key can\'t sign anything [1, 2]. The draft calls a secret shared with every site "impractical to scale" and "insecure, as key rotation is challenging" [1].') },
+ { title: 'It doesn\'t cover the crawlers in your licensing talks.', text: h(`GPTBot, ClaudeBot, PerplexityBot and Googlebot don't sign [13, 14, 15, 16]. Hidden crawlers stay hidden too. For those, robots.txt, IP lists and contracts remain the tools. ${a('/resources/ai-crawlers/', 'How to control which bots get in →')}`) },
+];
+
+const SIGNERS = [
+ ['OpenAI', `ChatGPT Work's Cloud browser, as ${c('https://chatgpt.com')}`, statusTag('live'), h('OpenAI\'s allowlisting guide; its key directory answers [17]')],
+ ['Google', `Some Google-Agent requests, as ${c('https://agent.bot.goog')}`, statusTag('experimental'), h('"We don\'t sign every request of a particular agent" [13]')],
+ ['Cloudflare', 'Its own fetchers, among them Browser Run, its crawler and AI Search', statusTag('live'), h('Cloudflare\'s docs; Cloudflare Radar\'s bot directory [18, 19]')],
+ ['Exa', 'ExaSearchBot and ExaUser', statusTag('live'), h('"Every ExaSearchBot request is cryptographically signed"; ExaUser is listed on Radar [19, 20]')],
+ ['Browserbase', 'Hosted browsers, for accounts Browserbase enables', statusTag('closed-beta'), h('"Contact Browserbase to enable Web Bot Auth for your account" [21]')],
+ ['Shopify', 'Cloudsurf, its agent browser for Sidekick', statusTag('live'), h('Listed on Radar, with a key directory that answers [19]')],
+ ['Amazon', 'Bedrock AgentCore Browser, when the customer turns it on', statusTag('preview'), h('"The feature is disabled by default and must be explicitly enabled" [22]')],
+];
+const NON_SIGNERS = h('The rest don\'t sign, the large training and search crawlers among them. OpenAI\'s GPTBot, OAI-SearchBot and ChatGPT-User, Googlebot, Anthropic\'s ClaudeBot, Claude-User, Claude-SearchBot and its hosted browser Claude-Agent, and Perplexity\'s PerplexityBot and Perplexity-User identify themselves by user agent, reverse DNS and published IP lists [13, 14, 15, 16, 19, 23, 24].');
+
+const KEY_FACTS = [
+ { value: '2.3%', text: h('Of the agent sessions Arkose Labs judged legitimate carried a verified signature, on login, registration and checkout pages, Q3 2026. Arkose\'s own data [26]') },
+ { value: '78.4%', text: h('Of ChatGPT agent\'s sessions were signed in the same report, "the only agent that signed at scale" [26]') },
+ { value: '12 of 41', text: h('Bots in Cloudflare Radar\'s AI crawler category can sign. GPTBot, ClaudeBot, Amazonbot and Meta\'s crawlers aren\'t among them [19]') },
+];
+
+const CAUTIONS = [
+ { text: h('A taken-over domain passes. Resolving the directory shows that "Whoever controls that URL says this key signs for it". The draft adds: "It does not say that the operator of that URL is honest" [1].') },
+ { text: h(`A registry speaks for many agents. A registry can give each agent its own subdomain and key directory, such as ${c('agent-x.registry.example')}, so each agent keeps an identity of its own and sites can tell them apart [31]. The registry holds or publishes all of those keys, so a breach there hits every agent at once: "a compromised registry could already impersonate every agent in its custody" [8]. FORA's Identity Service is such a registry [8].`) },
+];
+
+const ASK_CDN = [
+ { title: 'Do you verify Web Bot Auth signatures, and whose?', text: h('Cloudflare verifies bots that registered their key directory with it [12]. AWS WAF labels each signed request verified, invalid, expired or unknown [27]. Akamai, HUMAN and Vercel verify ChatGPT\'s cloud browser, by OpenAI\'s guide [28], and DataDome verifies bots that apply through its form [29].') },
+ { title: 'Can our rules treat signed, unsigned and failed requests differently?', text: h('AWS WAF lets a custom rule act on its label [27]. Shopify, for one, gives bots that don\'t sign "the strictest limits" [30].') },
+ { title: 'Will our bot rules let FORA buyers through?', text: h('Cloudflare\'s built-in check reads the format used before the working group adopted the draft, and rejects the adopted format FORA uses. On Cloudflare, FORA\'s own edge function does the check [12].') },
+];
+const ASK_TEAM = [
+ { title: 'Which of our pages are for sale?', text: 'Only those need a signature check. Open pages stay open.' },
+ { title: 'Where would FORA\'s edge function run?', text: h('It\'s a drop-in on Cloudflare Workers, Fastly Compute and Lambda@Edge [7].') },
+];
+
+const FORA_STEPS = [
+ { text: h('The identity is the domain. An agent is the host that serves its key directory: "If you control the domain, you control the key." A subdomain is a different party [5].') },
+ { text: h('Registration carries no password or API key, because the signature is the credential. The Exchange reads who is registering from the verified signature alone [9].') },
+ { text: h('Each signature covers more than the minimum: the method, the full address, a digest of the body and a one-time random value. It lives a few minutes. A request that fails the check is rejected, with "No fallback, no degraded mode" [5].') },
+ { text: h('The account belongs to the domain, and a key rotation keeps it [5].') },
+ { text: h('The signed link names the buyer\'s key, and FORA\'s edge function on your CDN checks it [7].') },
+];
+
+const ENG_SIGN = [
+ h("Send Signature-Agent as a dictionary and cover your member: (\"@authority\" \"signature-agent\";key=\"sig1\"). Draft -01 says \"Signers MUST send the dictionary form\" [1]."),
+ h("The member's value changed between drafts. In -00 it's an origin, and the verifier looks up the well-known path there. In -01 it's the URL of your key set, and a verifier \"fetches that URL directly\" [1, 4]."),
+ h("Cloudflare's verifier accepts the earlier quoted-string form, \"https://signature-agent.test\". It fails the dictionary form and the key component parameter [12]. Draft -01 lets a verifier accept the string form but doesn't require it [1]."),
+ h("Google sends g=\"https://agent.bot.goog\", a dictionary naming its origin, which is the form of draft -00. OpenAI sends the string \"https://chatgpt.com\" [4, 13, 17]."),
+ h("Set tag=\"web-bot-auth\", created and expires, at most 24 hours apart by the draft's recommendation. Set keyid to the key's base64url JWK SHA-256 thumbprint (RFC 7638; RFC 8037 Appendix A.3 for Ed25519) [1]."),
+ h("Cover @authority or @target-uri. A signature over @authority alone \"verifies against any method, path, or body sent to that authority until it expires\" [1]."),
+ h("To narrow it, also cover @method and @path or @target-uri, and send and cover Content-Digest for a body. The draft sets no nonce rule. Cloudflare says \"there is currently no nonce validation\" and, of expires, \"A minute is often sufficient\" [1, 12]."),
+ h("Algorithms should be registered in the HTTP Message Signatures algorithm registry, and the draft's test vectors use Ed25519 and RSA-PSS with SHA-512. Cloudflare \"accepts all valid Ed25519 keys found in your key directory\" [1, 12]."),
+ h("Serve the directory over https as a JWK Set. At the well-known path its media type must be application/http-message-signatures-directory+json [1]."),
+ h("Sign the response once per key over \"@authority\";req and content-digest. The draft recommends it, and Cloudflare requires it before it registers a directory [1, 12]."),
+ h("Rotate in three steps: add the new key before you use it, keep the old one until it expires, then remove it, so \"the directory's cache lifetime bounds how long a removed key keeps verifying\". The draft \"defines no revocation\" [1]."),
+];
+const ENG_VERIFY = [
+ h("Fetch the member's URL over https, accept only a 200 and follow no redirects. Guard the fetch against server-side request forgery. Attach no policy to the URL until a key from it verifies the signature [1]."),
+ h("\"Key lookup MUST be keyed on the (URL, key) pair, not on the key alone\" [1]."),
+ h("Cache the directory under normal HTTP caching rules and refresh it when it goes stale. A failed fetch \"MUST NOT evict a cached entry\", and failed lookups should stay cached five minutes at most [1]."),
+ h("To ask for a signature, send Accept-Signature, usually with a 403. A signature or nonce used more often than your policy allows can get a 429. Whether to require nonces \"is deployment policy\" [1]."),
+];
+const ENG_FORA = [
+ h("FORA signs the profile of draft -00, the one the working group adopted on September 1, 2026. A -00 verifier checks a FORA signature without knowing FORA, and ignores the members FORA adds to its directory [4, 5]."),
+ h("Its Signature-Agent is a dictionary naming the signer's https origin, such as sig1=\"https://research.acme.com\", covered with the key parameter. Cloudflare's verifier rejects that form [5, 12]."),
+ h("On every FORA call the signature also covers @method, @target-uri, content-digest and authorization, sent empty when there's no token. It carries a fresh 64-byte random nonce [5]."),
+ h("The lifetime is \"a few minutes\", and the docs' example uses 300 seconds. A FORA verifier that keeps a replay store refuses a signature it has already accepted [5]."),
+ h("A FORA directory must be signed. Each key is Ed25519 and carries not_before and not_after. A FORA verifier follows only members of -00's default directory type, so it ignores a jwks_uri or cimd member [5]."),
+ h("A directory may name a revocation_url. FORA verifiers poll it every 300 seconds, with 10% jitter, and each list revokes only its own keys [5]."),
+ h("In Web Bot Auth the identity is the directory's URL. FORA narrows it to the host, and a subdomain is a different party [5]."),
+ h("A signed link names the thumbprint of the buyer's key. The buyer sends its public key with the fetch, so the edge function checks the link, the thumbprint and the signature without fetching the buyer's directory [7]."),
+ h("An agent on its own domain holds its own key. On FORA's Identity Service, the service creates the key, \"keeps the private key in its key store\" and signs for the agent [8]."),
+ h("The docs name the cost of custody: \"a compromised registry could already impersonate every agent in its custody\" [8]."),
+ h("The Identity Service gives \"One developer, one subdomain, one key\". The subdomain is random and \"leaks no identity\", and it's the same at every Exchange, so Exchanges can link one developer's requests across them [1, 8]."),
+];
+
+// Quoted verbatim from draft -01, with its RFC 8792 line wrapping.
+const SIGNED_REQUEST = String.raw`NOTE: '\' line wrapping per RFC 8792
+
+Signature-Agent: sig2="https://signature-agent.test"
+Signature-Input: sig2=("@authority" "signature-agent";key="sig2")\
+ ;created=1735689600\
+ ;keyid="poqkLGiymh_W0uP6PZFw-dvez3QJT5SolqXBCW38r0U"\
+ ;alg="ed25519"\
+ ;expires=4889289600\
+ ;nonce="n9p433xm+NJ3ph3upfBIGmsuwHw387YV7Q/F+6BSpGCVjYCqQw6rznNA8PVVLySrAWsv0hQtFioQb6E1YsauiA=="\
+ ;tag="web-bot-auth"
+Signature: sig2=:CGokAoY/FrPpyuzTe1Fb2kLTAH9Idw6qSFd1+D77r1firsMRCxeuLu5e6O+ek3zQgDMAzjiXtLVfqE00xnABCQ==:`;
+
+const DIRECTORY = `GET /.well-known/http-message-signatures-directory HTTP/1.1
+Host: example.com
+Accept: application/http-message-signatures-directory+json
+
+HTTP/1.1 200 OK
+Content-Type: application/http-message-signatures-directory+json
+Cache-Control: max-age=86400
+{
+ "keys": [{
+ "kty": "OKP",
+ "crv": "Ed25519",
+ "kid": "NFcWBst6DXG-N35nHdzMrioWntdzNZghQSkjHNMMSjw",
+ "x": "JrQLj5P_89iXES9-vFgrIy29clF9CC_oPPsw3c5D0bs",
+ "use": "sig",
+ "nbf": 1712793600,
+ "exp": 1715385600
+ }]
+}`;
+
+const FAQ = [
+ { q: 'Do I need Cloudflare, or any one vendor, to check signatures?', a: h('No. The key directory sits on the agent\'s domain, and any site or vendor can check it by the draft\'s rules [1]. Each verifier decides which agents it knows. Cloudflare\'s checks the directories registered with it [12].') },
+ { q: 'Why not use OAuth instead?', a: h(`Use OAuth for what the agent may do. Web Bot Auth says which website is asking and carries no permissions; an OAuth token carries them [1, 43]. ${a(ID, 'How the two fit together →')}`) },
+ { q: 'Does requiring signatures affect search?', a: h('Googlebot doesn\'t sign, and Google tells sites to keep relying on IP addresses, reverse DNS and user agents alongside Web Bot Auth [13, 19]. FORA\'s edge never refuses Googlebot, Bingbot or Applebot [11].') },
+ { q: 'Will a signature rule refuse agents I want?', a: h('It can. Google signs only some Google-Agent requests and tells sites to fall back to the older checks [13]. On Arkose\'s network, most legitimate agent sessions carried no verified signature [26].') },
+ { q: 'Does Web Bot Auth track my readers?', a: h('No. It identifies agents, and a person reading your site signs nothing. One identifier links an agent\'s requests across sites, so the draft calls it unsuited to an identifier tied to one person [1].') },
+ { q: 'Does an agent have to register with each verifier?', a: h('Not by the draft, which asks for no prior relationship [1]. Cloudflare counts signatures from bots that registered their key directory through its bot submission form [12]. DataDome asks operators to apply through a form [29]. AWS WAF checks the signature against the agent\'s key directory and labels the result [27]. A FORA Exchange checks a signature from any key directory that meets FORA\'s rules, and an account takes one signed request [5, 9].') },
+];
+
+// The SourceList, numbered as the copy numbers it: an entry's position,
+// counted across the groups, is the N in a [N] marker. Generated from the
+// page copy's source list.
+const CHECKED = '2026-11-16';
+const src = (title: string, publisher: string, href: string) => ({ title, publisher, href, checked: CHECKED });
+const SOURCES = [
+ {
+ name: "The draft and the standard",
+ items: [
+ src("HTTP Message Signatures for automated traffic, draft-ietf-webbotauth-httpsig-protocol-01", "IETF webbotauth working group", "https://www.ietf.org/archive/id/draft-ietf-webbotauth-httpsig-protocol-01.txt"),
+ src("RFC 9421: HTTP Message Signatures", "IETF", "https://www.rfc-editor.org/rfc/rfc9421"),
+ src("Call for adoption of draft-meunier-webbotauth-httpsig-protocol, and the chairs' decision", "IETF web-bot-auth mailing list", "https://mailarchive.ietf.org/arch/msg/web-bot-auth/IHIsC9v_WDK_fv2SalA4KeMl8GI/"),
+ src("HTTP Message Signatures for automated traffic, draft-ietf-webbotauth-httpsig-protocol-00", "IETF webbotauth working group", "https://www.ietf.org/archive/id/draft-ietf-webbotauth-httpsig-protocol-00.txt"),
+ ],
+ },
+ {
+ name: "FORA",
+ items: [
+ src("Authentication", "FORA documentation", "/protocol/authentication/"),
+ src("Licensing terms", "FORA documentation", "/protocol/licensing-terms/"),
+ src("Signed URL verification", "FORA documentation", "/components/edge-function/signed-url-verification/"),
+ src("Identity Service", "FORA documentation", "/components/identity/overview/"),
+ src("fora.proto reference: RegisterRequest, RegisterResponse", "FORA documentation", "/reference/proto-fora/"),
+ src("Exchange manifest", "FORA documentation", "/protocol/exchange-manifest/"),
+ src("Bot detection", "FORA documentation", "/components/edge-function/bot-detection/"),
+ ],
+ },
+ {
+ name: "Signers",
+ items: [
+ src("Web Bot Auth (bot verification)", "Cloudflare docs", "https://developers.cloudflare.com/bots/reference/bot-verification/web-bot-auth/"),
+ src("Authenticate requests with Web Bot Auth (experimental)", "Google for Developers", "https://developers.google.com/crawling/docs/crawlers-fetchers/web-bot-auth"),
+ src("Overview of OpenAI crawlers", "OpenAI", "https://developers.openai.com/api/docs/bots"),
+ { title: "Does Anthropic crawl data from the web, and how can site owners block the crawler?", publisher: "Anthropic", where: "Claude help center, article 8896518", checked: CHECKED },
+ src("Perplexity crawlers", "Perplexity docs", "https://docs.perplexity.ai/docs/resources/perplexity-crawlers"),
+ src("ChatGPT Work's Cloud browser allowlisting", "OpenAI Help Center", "https://help.openai.com/en/articles/11845367-chatgpt-works-cloud-browser-allowlisting"),
+ src("Automatic request headers", "Cloudflare Browser Run docs", "https://developers.cloudflare.com/browser-run/reference/automatic-request-headers/"),
+ src("Bots directory", "Cloudflare Radar", "https://radar.cloudflare.com/bots/directory"),
+ src("Exa crawler", "Exa", "https://crawler.exa.ai/"),
+ src("Web Bot Auth", "Browserbase docs", "https://docs.browserbase.com/platform/identity/web-bot-auth"),
+ src("Web Bot Auth for AgentCore Browser", "AWS docs", "https://docs.aws.amazon.com/bedrock-agentcore/latest/devguide/browser-web-bot-auth.html"),
+ { title: "How do Anthropic-hosted browsers identify themselves?", publisher: "Anthropic", where: "Claude help center, article 18234133", checked: CHECKED },
+ src("SAIL, draft-kelley-webbotauth-sail-00", "IETF, individual Internet-Draft", "https://datatracker.ietf.org/doc/draft-kelley-webbotauth-sail/"),
+ src("Verified bots", "Cloudflare docs", "https://developers.cloudflare.com/bots/concepts/bot/verified-bots/"),
+ ],
+ },
+ {
+ name: "Numbers and verifiers",
+ items: [
+ src("The State of Agentic AI Trust Report, Q3 2026", "Arkose Labs", "https://www.arkoselabs.com/agentic-ai-trust-report-q3-2026"),
+ src("AWS WAF Bot Control rule group", "AWS docs", "https://docs.aws.amazon.com/waf/latest/developerguide/aws-managed-rule-groups-bot.html"),
+ src("Help ensure cloud browser access to your site", "OpenAI for developers", "https://developers.openai.com/commerce/agent-access"),
+ src("Bot authentication", "DataDome docs", "https://docs.datadome.co/docs/bot-authentication"),
+ src("Bots and agents should identify themselves via Web Bot Auth", "Shopify developer changelog", "https://shopify.dev/changelog/bots-and-agents-should-identify-themselves-via-web-bot-auth"),
+ ],
+ },
+ {
+ name: "Allow lists and registries",
+ items: [
+ src("Hosted Key Directories for Web Bot Auth, draft-singh-webbotauth-hosted-directories-00", "IETF, individual Internet-Draft", "https://datatracker.ietf.org/doc/draft-singh-webbotauth-hosted-directories/"),
+ src("AI Crawler and Agent access policies", "Cloudflare docs", "https://developers.cloudflare.com/bots/additional-configurations/block-ai-bots/"),
+ src("Signature Agent Card and registry, draft-meunier-webbotauth-registry-04", "IETF, individual Internet-Draft", "https://www.ietf.org/archive/id/draft-meunier-webbotauth-registry-04.txt"),
+ ],
+ },
+ {
+ name: "The debate",
+ items: [
+ src("WebBotAuth Direction", "the working group's chairs, IETF web-bot-auth mailing list", "https://mailarchive.ietf.org/arch/msg/web-bot-auth/gxkMGMC75Sm1LO9WPvZbPKYhNAc/"),
+ src("Minutes, IETF 126 webbotauth session", "IETF", "https://datatracker.ietf.org/doc/minutes-126-webbotauth/"),
+ src("Re: Harms", "Mark Nottingham, IETF web-bot-auth mailing list", "https://mailarchive.ietf.org/arch/msg/web-bot-auth/5wm6ZdVbl-hb48_GE7XCnJaRICE/"),
+ src("The Free and Open Web Is Under Attack at the IETF", "EFF", "https://www.eff.org/deeplinks/2026/06/free-and-open-web-under-attack-ietf"),
+ src("Re: Harms", "Mirja Kühlewind, IETF web-bot-auth mailing list", "https://mailarchive.ietf.org/arch/msg/web-bot-auth/M7Gpx_Vow1pCJ65vt2MAvPsqtPI/"),
+ src("Moving past bots vs. humans", "Cloudflare blog", "https://blog.cloudflare.com/past-bots-and-humans/"),
+ src("New York becomes first state to ban bots that scrape news sites", "Bloomberg Government", "https://news.bgov.com/bloomberg-government-news/new-york-becomes-first-state-to-ban-bots-that-scrape-news-sites"),
+ src("NY State Senate passes stealth crawler bill", "News/Media Alliance", "https://www.newsmediaalliance.org/ny-state-senate-stealth-crawler-bill-passage/"),
+ src("EFF joins 18 civil rights organizations calling on Governor Hochul to reject stealth crawler bill", "EFF", "https://www.eff.org/deeplinks/2026/08/eff-joins-18-civil-rights-organizations-calling-governor-hochul-reject-stealth"),
+ ],
+ },
+ {
+ name: "Identity and authorization",
+ items: [
+ src("RFC 6749: The OAuth 2.0 Authorization Framework", "IETF", "https://www.rfc-editor.org/rfc/rfc6749"),
+ ],
+ },
+];
+
+const MORE = [
+ { href: ID, text: 'How AI agents prove who they are, and what they may do' },
+ { href: '/resources/standards-map/', text: 'The standards map' },
+ { href: '/resources/ai-crawlers/', text: 'How to control which bots get in' },
+ { href: '/resources/content-marketplaces/', text: 'Content marketplaces and FORA' },
+];
+---
+
+
+ How it works
+ What it changes for you
+ Building an agent? The detail for engineers →
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
{TAKEAWAY}
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Ask your CDN or bot vendor
+
+
Ask your engineers
+
+
+
+
+
+
+
+
Your readers and search crawlers reach your open pages as before. What follows is only for the pages you sell access to.
+
+
+
+
+
+
+
+
A deny list is weaker: it stops one domain, and domains are cheap. The draft agrees:
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
If you sign
+
{ENG_SIGN.map((t) => )}
+
+
{SIGNED_REQUEST}
+
+
{DIRECTORY}
+
If you verify
+
{ENG_VERIFY.map((t) => )}
+
What FORA signs
+
{ENG_FORA.map((t) => )}
+
+
+
+
+
+
+
+
+
+
+
+
Sources
+
+ All sources
+
+
+
Sources are checked on the date of publication.
+
+
+
More in Resources
+
+
+
+
+
+ See your offers
+ How FORA agents sign
+
+
+
+
+
+
diff --git a/website/src/styles/landing.css b/website/src/styles/landing.css
index 77171dda..135a0d6b 100644
--- a/website/src/styles/landing.css
+++ b/website/src/styles/landing.css
@@ -191,6 +191,57 @@
outline-offset: 5px;
}
+/* ========= RESOURCE-PAGE ATOMS =========
+ Inline pieces that also land inside HTML strings (a GlanceCard cell, a
+ table cell, a step), so their styles can't be scoped to one component.
+ The markup comes from src/lib/resource-markup.ts. Both read the Band's
+ --lp-* variables, so they work on a black, near or light band. */
+
+/* Status tag: monochrome on purpose. The states differ by fill and line, so
+ they read on every band and never compete with FORA's blue. Live is filled;
+ Beta, Closed beta, Pilot and Preview are outlined; Experimental has a
+ dashed dot; Design and Roadmap are dashed;
+ Not public is dotted and muted. Every icon box is the same size, so no
+ pill drops below the text line. */
+.status-tag {
+ display: inline-flex;
+ align-items: center;
+ gap: 0.4rem;
+ vertical-align: middle;
+ position: relative;
+ top: -0.08em;
+ font-family: var(--mono);
+ font-size: 0.6875rem;
+ font-weight: 500;
+ line-height: 1;
+ letter-spacing: 0.08em;
+ text-transform: uppercase;
+ white-space: nowrap;
+ padding: 0.3rem 0.6rem 0.3rem 0.5rem;
+ border-radius: 999px;
+ border: 1px solid var(--lp-heading);
+ color: var(--lp-heading);
+}
+.status-tag::before { content: ''; width: 0.5rem; height: 0.5rem; border-radius: 50%; flex: none; }
+.status-tag--live { background: var(--lp-heading); color: var(--lp-bg); }
+.status-tag--live::before { background: var(--lp-bg); }
+.status-tag--beta::before,
+.status-tag--closed-beta::before { background: linear-gradient(90deg, var(--lp-heading) 50%, transparent 50%); border: 1px solid var(--lp-heading); }
+.status-tag--pilot::before,
+.status-tag--preview::before { border: 1px solid var(--lp-heading); }
+.status-tag--experimental::before { border: 1px dashed var(--lp-heading); }
+.status-tag--design,
+.status-tag--roadmap { border-style: dashed; }
+.status-tag--design::before { border: 1px dashed var(--lp-heading); }
+.status-tag--roadmap::before { border-radius: 0; background: linear-gradient(var(--lp-heading), var(--lp-heading)) center / 100% 1px no-repeat, linear-gradient(var(--lp-heading), var(--lp-heading)) right center / 1px 100% no-repeat; }
+.status-tag--not-public { border-style: dotted; border-color: var(--lp-muted); color: var(--lp-text); }
+.status-tag--not-public::before { border-radius: 0; background: linear-gradient(var(--lp-muted), var(--lp-muted)) center / 100% 1px no-repeat; }
+
+/* Citation marker: "[1, 2]", each number a link to the page's SourceList. */
+.cite { white-space: nowrap; font-size: 0.8125em; color: var(--lp-muted); font-variant-numeric: tabular-nums; }
+.cite a { color: var(--lp-link); text-decoration: none; }
+.cite a:hover { text-decoration: underline; }
+
@media (max-width: 768px) {
.hero-content { padding: 3rem 1.5rem; }