What it does
An MCP server that records what an AI agent was authorized to do and what it actually did. Tool calls produce signed policy decisions and evidence chains stored in a SQLite ledger, which a third party can verify offline using the exported evidence pack and a public key.
Use cases
- 01Prove an agent stayed within its agreed scope
- 02Export a signed evidence pack for a client
- 03Audit agent tool calls from a local ledger