Fava Trails

[](https://pypi.org/project/fava-trails/) [](LICENSE) [](https://github.com/MachineWisdomAI/fava-trails/actions) [](https://pypi.org/project/fava-trails/) [](https://github.com/MachineWisdomAI/fava-trails)

FAVA Trails

**Federated Agents Versioned Audit Trail** — Git-native, curated memory for AI agents via MCP.

Every thought, decision, and observation is stored as a markdown file with YAML frontmatter in a Git repo you control, with crash-proof persistence and a versioned audit trail. Agents interact through [MCP](https://modelcontextprotocol.io/) tools — they never see VCS commands.

Governed recall

FAVA is the governed institutional record for decisions, observations, validation, and lineage. It is not the operational working-context store. Default `recall` and `get_thought` expose current approved records only. Explicit `mode="authoring"` retrieves only the server-configured agent's draft/proposed records; operator-only `mode="history"` selects lifecycle statuses and superseded records. Neither a namespace nor a supplied `agent_id` grants access. See [governed-recall.md](docs/governed-recall.md) for identity setup, compatibility, approval provenance, and interrupted-write recovery.

The operator configures `FAVA_TRAILS_AGENT_ID` on a dedicated process; caller `agent_id` must match it. A shared endpoint is one identity boundary. Configure `FAVA_TRAILS_OPERATOR=1` only on a separate operator-controlled endpoint.

Why

• Supersession tracking — a proposed correction leaves the original current; approved replacements make predecessors historical. No