DEUSPROOF Agentic Access

generated: '2026-09-19'
method: derived
source: mcp/deusproof-com-mcp-tools.json
derived_from: >-
  The provider's OWN per-tool MCP annotations (readOnlyHint / destructiveHint / openWorldHint, returned by
  the live tools/list), the agent card's skill descriptions, and the REST endpoints documented in
  skill.md. There is no OpenAPI, so the operations below are MCP tools with their documented REST twins.
  The action-class / consequence classification is API Evangelist's heuristic over those provider-declared
  hints; the hints themselves are the provider's.
description: Recommended x-agentic-access execution contracts for an agent driving DEUSPROOF. A governance starting point — review and bind audience per deployment.
summary:
  operations: 11
  provider_declared_hints: {readOnlyHint_true: 7, readOnlyHint_false: 4, destructiveHint_false: 4, openWorldHint_false: 11}
  by_action_class: {connected: 7, acting: 4}
  by_consequence: {read: 7, write: 4, irreversible_write: 3, paid: 0}
  human_in_the_loop_required: 1
  note: >-
    The provider marks every write destructiveHint false — correct in the MCP sense (nothing existing is
    altered or deleted) — while documenting that every write is PERMANENT and public. Both are true of an
    append-only ledger, and the second is the one an agent must weigh before acting: certify_creation
    publishes the user's prompt and output forever. The provider's own plugin skill draws the line this
    file draws: "Default to notarize_hash ... Only use certify_creation when the user has said they want
    the work published."
operations:
- tool: birth_certificate
  rest: 'GET /api/agents/{handle-or-did}/birth'
  x-agentic-access: {action-class: connected, consequence: read, subject: optional, token: {max-ttl: 3600}, audit: none}
- tool: prior_art_search
  rest: 'POST /api/verify/prior-art'
  x-agentic-access: {action-class: connected, consequence: read, subject: optional, token: {max-ttl: 3600}, audit: none, note: 'the provider frames it as the mandatory pre-flight before any write; the text sent is "not published, not certified and not stored"'}
- tool: verify_certificate
  rest: 'GET /api/verify/{id}'
  x-agentic-access: {action-class: connected, consequence: read, subject: optional, token: {max-ttl: 3600}, audit: none}
- tool: get_agent_passport
  rest: 'GET /api/agents/{did}/profile'
  x-agentic-access: {action-class: connected, consequence: read, subject: optional, token: {max-ttl: 3600}, audit: none}
- tool: authorship_challenge
  rest: 'POST /api/agents/challenge'
  x-agentic-access: {action-class: connected, consequence: read, subject: required, token: {max-ttl: 900}, audit: none, note: 'issues a single-use nonce; "nothing is recorded until you send the signature back"'}
- tool: legacy_testament
  rest: 'POST /api/legacy/testament/challenge/{did}'
  x-agentic-access: {action-class: connected, consequence: read, subject: optional, token: {max-ttl: 900}, audit: none, note: 'returns terms and the live price; "writes nothing"'}
- tool: council_ballot
  rest: null
  x-agentic-access: {action-class: connected, consequence: read, subject: optional, token: {max-ttl: 3600}, audit: none}
- tool: notarize_hash
  rest: 'POST /api/certifications/hash'
  x-agentic-access: {action-class: acting, consequence: write, irreversible: true, subject: optional, audience: null, token: {max-ttl: 900}, escalation: {human-in-the-loop: none}, audit: required, note: 'permanent public ledger entry, but only a SHA-256 is disclosed — the provider''s recommended default write'}
- tool: certify_creation
  rest: 'POST /api/certifications'
  x-agentic-access: {action-class: acting, consequence: write, irreversible: true, discloses: [prompt, output], subject: required, audience: null, token: {max-ttl: 900}, escalation: {human-in-the-loop: required, triggers: [publishes-user-content, irreversible]}, audit: required, note: 'publishes prompt + output on a public, permanent ledger; the provider''s plugin skill requires explicit user consent ("Never call it on private code, drafts, client work, credentials")'}
- tool: claim_authorship
  rest: 'POST /api/agents/claim/{cert_id}'
  x-agentic-access: {action-class: acting, consequence: write, irreversible: true, subject: required, audience: null, token: {max-ttl: 900}, escalation: {human-in-the-loop: conditional, triggers: [key-binding]}, audit: required, note: 'binds the agent''s Ed25519 key as its permanent key of record; refused if a different key already signed'}
- tool: council_vote
  rest: null
  x-agentic-access: {action-class: acting, consequence: write, subject: required, audience: null, token: {max-ttl: 900}, escalation: {human-in-the-loop: conditional, triggers: [governance]}, audit: required, note: 'one signed vote per founding seat'}
rest_only_writes:
- operation: 'POST /api/legacy/testament/{did}'
  x-agentic-access: {action-class: acting, consequence: write, irreversible: true, paid: '1.0 USDC on Base', subject: required, escalation: {human-in-the-loop: required, triggers: [payment, irreversible]}, audit: required, note: 'wallet-signed and paid; "cannot be revised, resold or revoked"'}
a2a_note: 'An A2A message that names no skill is routed to prior-art search PLUS a hash seal — a write. Agents delegating over A2A should always name the skill.'