Agent evidence infrastructure

AccordTrace

Cryptographic proof for AI-agent actions.

Create a tamper-evident record, pass it to another agent, and verify the evidence hash later. No account, call or manual approval is required for the anonymous free tier.

Accord Trace proves service attestation and integrity, not truth or legal validity.

POST /api/v1/proofs ANONYMOUS
{
  "data": {
    "event": "deployment.complete",
    "release": "2026.08.24"
  },
  "metadata": {
    "source": "agent-a"
  }
}
SHA-256 + ES256Raw data not stored

Discover → call → verify

One machine-readable path.

An agent discovers Accord Trace through A2A, MCP, OpenAPI or llms.txt and can create a proof immediately.

  1. 01DiscoverRead Agent Card or MCP tools
  2. 02SubmitSend content or SHA-256
  3. 03ReceiveGet proof ID and ES256 attestation
  4. 04VerifyCheck proof and optional content match

When to call

Preserve the checkpoint.

Use Accord Trace before handing work between agents, recording approval, preserving an artifact state, or completing an important autonomous action.

  • 01Agent actions and task completion
  • 02Agreements and human approvals
  • 03API outputs, documents and deployments
Accord Trace machine-readable proof response and verification result
Every proof includes verification instructions for the next agent.

Public interfaces

Built for machine discovery.

Security boundary

Integrity is not truth.

Accord Trace attests that a specific evidence hash was recorded by the service at a stated time. It does not prove that the underlying claim is true, complete, authorised, fair, legal, or observed by an independent timestamp authority.

Read the exact guarantees