MemoMind

Why we included this project

Anyone who runs Claude Code or another MCP-capable coding agent knows the frustration of re-explaining your architecture and naming conventions every time a session starts. MemoMind fixes that by keeping a persistent, structured memory on your own machine, with GPU-accelerated vector search backed by PostgreSQL and pgvector, so nothing leaves your hardware. It pulls facts out of conversations into a knowledge graph, letting the agent recall your tech-stack choices and past decisions instead of asking again. Because it speaks MCP over stdio, it drops into an existing agent setup without custom glue, and the import tools can bring in years of ChatGPT or Gemini history. If you are tired of your agent treating you like a stranger each morning, this is a straightforward way to make context stick while keeping the data private.