mnemaA tamper-evident audit trail for AI-agent work

You drive, agents execute — every change stamped with who authorized it and which agent ran it, in a log you can prove wasn't altered.

Client integration

Configuration reference

GitHub

Provable

Every mutation appends to a hash-chained, keyed, per-machine-signed audit log. `mnema doctor` detects edits, truncation, replays, deletion, and downgrade.

Human-in-the-loop

Agents move work through workflow gates that reject invalid transitions. You approve through the terminal; the agent can't route around you.

Local-first

Zero telemetry, no remote services. SQLite + plain-text Markdown/JSONL in your repo — the files outlive mnema and open in any editor.

This site is the reference for mnema. Start with Client integration to wire an agent to the rail, or jump to the Configuration reference for every config key.