Schema

Fields, grouped

Thirteen fields in four groups, followed by what the record declares. The fields and views are defined in the draft specification. Agent type takes one value per AIAO dimension, with cognitive capabilities listed as properties.

Parties around the agentThe developer builds the agent and the deployer operates it. The user authorises its use per mandate, under policy. Trust checks the agent's identity, mandate and policy. Developerwrites the agent Deployerruns the agent Agent name · AIAO typeprotocol · where it operatesmandate · authority · policy Userauthorises use Trustidentity · mandate · policy buildsoperates per mandate,under policychecks

Descriptions in italics (protocol, authority, policy) are proposals still open for discussion in the W3C AI KR Community Group.

Builder

Write a record

Fill in the form; the card on the right updates as you type. Add it to the register when done.

Identity
Interface
Mandate
Parties
Register

Agent trust register

The first row is an invented example to show the shape of a record. Delete it when adding your own.

The register is kept in this browser only. Copy the CSV or JSON to keep it elsewhere.

Protocols

Mappings

Which agent identity and trust specifications have an element matching each field of the trust record. A first reading of the specifications listed in the Agent Trust Protocols and Specifications Explorer, checked 18 November 2026. Each project owner is invited to confirm or correct their row.

GroupTrust record fieldSpecifications with a matching element
IdentityName or identifierARIA (did:aria), ATP (did:atp), ANP (did:wba), MCP-I, NTI-1, APS, ODIS (agent package and runtime instance), WIMSE
IdentityAgent typeNone of the specifications read. No specification types the agent by role, capability or embodiment; this is where AIAO adds to them.
InterfaceProtocolImplicit only: MCP-I and the China Mobile MCP draft bind to MCP; UCP is itself the protocol.
InterfaceWhere it livesARIA (DNS pointer to the Agent Identity Document), ATP (did:atp builds on did:wba: DNS and HTTPS resolution), ANP (web-based DID), Web Bot Auth (key directory URL)
InterfaceKey publisherAgent's own origin: Web Bot Auth, UCP, PAP, PACT. Issuer or network chosen by the verifier: KYAPay, Visa TAP, x402. Registry or domain: ARIA.
MandateMandate / functionsAAE (MANDATE block), AMEP (the delegated mandate), RTTP (names the intended action), ARIA (scopes)
MandateAuthorityODIS (delegated authority), MCP-I (delegation), APS (authority narrowed at each hop), NTI-1 (capabilities), AuthZEN (authorization decisions), ATP (mandates, in draft)
MandatePolicyAAE (CONSTRAINTS block), NTI-1 (caveats: expiry, value, path limits), ARIA (ARIA AP admission), AuthZEN (decision)
PartiesDeveloperNone of the specifications read.
PartiesDeployerARIA (organisation that answers for the agent), ODIS (principal)
PartiesUserODIS (delegated user identity), AMEP (the person who delegated), NTI-1 (principal-bound identity)
PartiesOwner, FunderNone of the specifications read.
TrustTrust assertionAAE (VALIDITY block), ATP (trust scoring and credentials, early draft), AMEP (each check returns preserved, violated or unresolved), IQA (whether attestation standing holds), ARIA (revocation status)

Closest matches

ARIA touches the most fields: identifier, location, keys, scope, admission, accountable organisation and validity.

AAE mirrors the Mandate group: its MANDATE, CONSTRAINTS and VALIDITY blocks line up with Mandate, Policy and the trust assertion.

ODIS asks who the agent acts for, what authority it was given and which instance is acting, matching User, Authority and Identity.

AMEP tests whether an agent stayed within its mandate, the question the trust assertion asks, and adds a third outcome, unresolved, for missing evidence.

Where the record goes further

Agent type, Developer, Owner and Funder appear in none of the specifications read. These fields carry the AIKR and AIAO contribution.

Where the specifications go further

Delegation chains (AMEP, ODIS, APS), revocation and validity windows (ARIA, AAE), and evidence of what an agent did (AMEP, APS, NTI-1). Candidate additions to the record: a delegation chain and a lifecycle field.

This is an editorial reading. Rows for ODIS, MCP-I and the AuthZEN agent drafts rest on summaries of the specifications, not their full text. Project owners can submit a mapping profile for their protocol (field map with spec section references, an example record from their use case, and evidence level) as a pull request or issue on w3c-cg/aikr/trust, or by posting to public-aikr@w3.org with [trustvisualiser] in the subject.

Acknowledgements

Contributions

Contributions to the visualiser and its draft specification. More are invited: open an issue or pull request in the w3c-cg/aikr trust folder, or post to public-aikr@w3.org with [trustvisualiser] in the subject.

ContributorContributionStatus
Larry Lewis (W3C Agent Trust Protocol CG)Review of the trust record model: separate declared, verified, permitted and observed states; declared functions from granted authority; identifiers from service endpoints; test the role mappings against deploymentsPartly implemented: "declared, not verified" labels and an "unknown" verification state in the visualiser. Other points open for discussion as issues
Yanfeng Liu (Nanjing University, V19)Reviewer for audit and accountability; reproducible examples for evidence freshness, human approval and revocationListed as reviewer; examples offered
Francesco Marinoni MorettoCommerce identity rows with pinned sources; who publishes the key set a verifier trusts (agent's own origin, or an issuer or network the verifier chose)Contributed: rows confirmed with corrections; key-publisher column proposed and added to the visualiser and draft as a proposed field; draft alignment issue between UCP and Web Bot Auth confirmed and reported to UCP (issue 1080)
Kenne Ives (AgentAvow)did:web example and signed tool-definition test vectors; pinning the key from a source independent of the issuerOffered; not yet run