FAVA Trails

FAVA Trails is an open-source memory system for AI agents that separates draft thoughts from approved shared knowledge. Agents save observations as drafts, submit them for review through a configurable gate, and recall approved records with full change history. Records are stored as Markdown files in a Git repository you control and exposed through the Model Context Protocol.

Designed for teams running multiple agents across sessions and environments, FAVA Trails prevents stale assumptions from spreading between agents by making memory curation explicit. The system preserves corrections in version history so later readers see what changed and why. Built with Python and Git-native storage, it integrates with MCP-compatible AI agents.