Agent Authorization & Lifecycle

Note (dated 2026-08-04): a chain composes independent computations performed in an analytical order. No step's output is carried into the next step by the execution model; each step computes from its own declared inputs. This is true of every chain on this site, including chains whose per-step copy below reads as a data handoff.

Reading mode (READINESS-BUILD): These checks can be gathered in any order; the sequence below is for convenience. Each step audits a separate MCP governance fact (tool-grant scope, on-behalf-of mandate, task-state legality) and none depends on another’s result.

Audit scoped and revocable MCP tool grants — each tool carries an explicit scope, a revocation endpoint is configured, and token rotation posture is healthy (art-150) → validate the agent on-behalf-of (OBO) mandate: subject, bounded scope, intent, and non-expired validity window; mismatch or expiry returns REFUSE (art-151) → validate that the long-running task state transitions are legal per the new MCP specification state machine (art-152). Exports lifecycle attestation with execution_hash. Zero network, zero PII.

OpenChainGraph · 3 Steps compliance mandate New MCP Spec 2026 OBO Mandate · Task Lifecycle W3C VC §13.11 Hash-Anchored §4 chain_depth:3 · Zero PII

Chain Stages · 3 Steps

1ROOT · D0node

MCP Tool Scope & Revocation Auditor art-150

Scope and revocation audit feeds OBO mandate validator

MCP Call · audit_mcp_tool_scope_revocation

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "audit_mcp_tool_scope_revocation",
    "arguments": {
      "tool_grants": [
        { "tool": "search_files", "scopes": ["files:read"] },
        { "tool": "write_file", "scopes": ["files:write"] }
      ],
      "revocation_endpoint": "https://auth.example.com/revoke",
      "token_created_unix": 1000000,
      "now_unix": 1001000,
      "max_token_age_s": 3600,
      "next_token_present": false
    }
  },
  "id": 1
}

↓

2MIDDLE · D1node

Agent On-Behalf-Of (OBO) Mandate Validator art-151

OBO mandate verdict feeds task lifecycle validator

MCP Call · validate_agent_obo_mandate

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "validate_agent_obo_mandate",
    "arguments": {
      "mandate": {
        "subject": "user:alice@example.com",
        "intent": "retrieve_invoice_data",
        "scope": ["invoices:read", "payments:read"],
        "valid_until_unix": 9999999999
      },
      "now_unix": 1000000000
    }
  },
  "id": 2
}

↓

3TERMINAL · D2node

MCP Task Lifecycle State Machine Validator art-152

Exports task lifecycle validity with execution_hash — final stage

MCP Call · validate_mcp_task_lifecycle

{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "validate_mcp_task_lifecycle",
    "arguments": {
      "transitions": [
        { "from": "working", "to": "input_required" },
        { "from": "input_required", "to": "working" },
        { "from": "working", "to": "completed" }
      ]
    }
  },
  "id": 3
}

Question: The agent presents an on-behalf-of mandate. Would the validator still accept it if the clock passed the validity window? Run the AINumbers MCP tool `run_chain` with {"chain":"agent-authorization-lifecycle"} and read `art-151-agent-obo-mandate-validator.verdict`. Re-run with `inputs` for `art-151-agent-obo-mandate-validator`, reusing the values that result echoed, with `now_unix` set to 4102444800. Compare the same field. Verify: call `verify_execution_hash` (https://mcp.ainumbers.co/mcp) with `claimed_hash` set to `composite_execution_hash` and the full `composite_artifact`. Ledger, for a human re-check: https://ledger.ainumbers.co/ PII rule: send synthetic or anonymised inputs only. The MCP server runs these kernels and logs no payloads. Chain page: https://ainumbers.co/chaingraph/chains/agent-authorization-lifecycle.html

Agent Authorization & Lifecycle

Ask your agent