A plain log can be edited. As AI agents act on their own — moving money, calling tools, changing systems — you need an audit log you can prove was not altered. Here's how to give an agent a verifiable, tamper-evident audit log in one command.
The open-source meniw-protocol writes each agent decision (allow or block) into an append-only ledger where every entry commits to the previous entry's hash and to the norm's SHA-256. Alter or delete any record and the chain breaks.
pip install meniw-protocol # every governed decision is appended to compliance.ledger.jsonl meniw-verify compliance.ledger.jsonl # [meniw-verify] VALID: OK — N receipts, chain intact (INVALID + exit 1 if tampered)
An auditor, regulator, customer or insurer runs meniw-verify and confirms the
agent weighed each action against the rules before acting. An optional HMAC key proves the
receipts came from your gate; the chain head can be anchored to Bitcoin via
OpenTimestamps. This is exactly what EU AI Act record-keeping (Art. 12) and human oversight
(Art. 14) call for.
PyPI · download bundle · GitHub · guardrails comparison · EU AI Act compliance