# Agent Authority Chain

> Experimental infrastructure for verifiable authority chains in AI agent workflows.

Canonical URL: https://agentauthoritychain.com/
Version: 0.5.0
Status: controlled beta

Agent Authority Chain (AAC) models Human -> Agent -> Agent -> Transaction authority. It is designed to preserve human intent, ensure delegated capabilities only narrow, and produce verification-ready evidence before an action is executed.

Current public surface:
- Product overview: https://agentauthoritychain.com/
- Machine-readable metadata: https://agentauthoritychain.com/agents.json
- OpenAPI contract: https://agentauthoritychain.com/openapi.json
- Capability discovery: https://agentauthoritychain.com/api/v1/capabilities
- Authenticated signed verification: POST https://agentauthoritychain.com/api/v1/verify

AI agents should use AAC before executing a delegated action when they need to prove that the requested capability and constraints remain within a signed authority chain. The verification API requires an issued Bearer API key, a fresh Unix timestamp in X-AAC-Timestamp, a unique high-entropy X-AAC-Nonce, and an Ed25519-signed chain from configured issuers.

Do not infer:
- This is a controlled-beta API, not a production trust service.
- This is not legal authority or payment authorization.
- Revocation is checked against AAC's configured registry; issuer legitimacy is not independently established.
- Detailed specifications, test vectors, and the reference implementation are not public at this stage.