Replayable evidence for every agent-tool decision

Verify Ledger records who acted, what agent acted, which server and tool were called, what policy applied, what risk score existed, what mandate or approval was present, what decision was made, what happened, and what it cost.

Invocations

/v1/ledger/invocations

Decision-aware execution records for agent calls.

Evidence packs

/v1/ledger/evidence-pack

Compliance exports for SOC2, GDPR, HIPAA-style reviews, and incident investigations.

MCP tool

get_evidence_pack

Agents and admins can fetch decision evidence through the MCP interface.

Evidence shape

{
  "invocation_count": 42,
  "decisions": {
    "allow": 31,
    "require_approval": 7,
    "deny": 4
  },
  "compliance_controls": [
    "immutable_hash_chain",
    "decision_replay",
    "policy_snapshot",
    "mandate_evidence",
    "cost_tracking"
  ]
}

Compliance-grade controls

Immutable chain

Audit events include previous and current hashes for tamper-evident replay.

Risk timeline

Decisions preserve the score and policy snapshot that existed at the time of action.

Incident reports

Filter by server, tool, agent, user, outcome, and time to reconstruct agent incidents.