record_stateWhat the governed record currently stores and exposes.
AI INTEGRATION
AI should not have to guess what a provenance claim means.
Governed records expose record state, lifecycle state, trust state, actionability, and authority state as distinct machine-readable fields. An AI can query, preserve, and relay each field without collapsing them into a single score or boolean.
Governed recordsFail-closed authorityLifecycle continuity
Do not collapse these fields.
lifecycle_stateWhether the record is active, suspended, reactivated, superseded, revoked, or expired.
trust_stateWhat the evidence and policy determination support, including limitations.
actionabilityWhether a specific purpose-bound action is allowed, denied, qualified, or unknown.
authority_stateWhether a separate authority such as signing, authoritative issuance, marks, or registry publication is actually activated.
What an AI agent can state, must qualify, and must not infer.
Record state, trust tier, lifecycle status, allowed claim scope, and known limitations — all as-of a specific timestamp.
Evidence scope, correspondence limits, submitter classification, remaining unknowns, and any purpose-bound restriction on actionability.
That a stored status implies authoritative issuance, signing authority, mark authorization, or registry authority unless the live authority state independently confirms activation.
GET /api/v1/trust/{publicId}/human — public, no API key required
The public human BFF endpoint resolves a governed record without an API key. It returns the identical canonical MachineTrustResponse via a server-side proxy. The machine endpoint (/machine) requires a Bearer API key with trust:read scope. A returned record state does not mean authoritative issuance, signing, certification-mark authority, or public registry authority is active. authority.issuance_authorized=false and authority_state=NOT_AUTHORIZED in all current responses.
GET /api/v1/trust/DET-V1-0CC5D81A36B661A7E9DBE0D2/human // No Authorization header required. Returns MachineTrustResponse. // authority.issuance_authorized=false — issuance is NOT_AUTHORIZED. // Machine API (requires Bearer token): GET /api/v1/trust/DET-V1-0CC5D81A36B661A7E9DBE0D2/machine Authorization: Bearer pv_live_xxxxRead the quickstart →
Use the governed tool contract when an agent needs an action, not only a lookup.
The MCP surface exposes governed tools for machine consumers. Each MachineTrustResponse includes prohibited_inferences — a machine-readable list of inferences an AI must not make from the current record state. Boundary enforcement is per-tier and per-purpose. Production MCP runtime requires the applicable authorized integration; the public machine-trust lookup remains distinct from privileged operations.
POST /api/v1/mcp
Content-Type: application/json
Authorization: Bearer pv_live_xxxx
{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"resolve_trust","arguments":{"subject_public_id":"DET-V1-..."}}}Record state is not authority state.
An AI should preserve evidence scope and limitations, check lifecycle, evaluate purpose-bound actionability where needed, and never convert a stored status such as “issued” into a claim that the record was authoritatively issued unless the live authority state independently proves that activation.
Request production access ↗PUBLIC TRUST
Trust remains inspectable.
Governed records expose their evidence scope, trust state, lifecycle state, limitations, and authority boundary without upgrading record state into authority that is not live.