Agent Memory

Stores AI coding agent memory as Markdown files in your git repo instead of a cloud database or vendor service. Exposes three MCP tools: fetch_context retrieves relevant facts via FTS, propose_update stages structured changes to memory sections, and memory.status shows what's pending. Updates stage for human review before they land, so you can diff exactly what the agent wants to commit. Includes secret and PII scanning on every write. Ships with adapters for Claude Code, Cursor, Gemini CLI, and AGENTS.md/Codex workflows. The git merge driver handles concurrent edits by merging at the section level rather than triggering text conflicts. Reach for this when you want durable agent knowledge that lives in version control, survives across sessions, and stays reviewable by the team.