memory-mcp

Local Brain MCP

Local Brain MCP is a small MCP server that you run locally on your machine. It stores and retrieves your personal memories in a local SQLite database and is intended to be accessed over HTTP by MCP‑aware clients such as Cursor (e.g. at http://localhost:3000/mcp).

Goals

• Run locally — start the server on your own machine; no cloud deployment required.
• Keep memory data on your own disk (memory.db or a custom path).
• Optionally expose the same memory from multiple devices via a tunnel (e.g. Cloudflare Tunnel) if you choose.
• Provide a simple, transparent API for saving and querying memories.

High‑level architecture

• MCP Server: Python + FastMCP (server.py).
• Storage: SQLite database file (memory.db by default).
• Tools:
  • save_memory — insert a memory row.
  • update_memory — patch fields on an existing row.
  • delete_memory — delete a row by id.
  • fetch_memories — RAG-style search (semantic by default, keyword fallback) or recent list when query is empty.