Skip to content
Back to projects

Apr 2026

Mnemosyne

Long-term vector memory for agents and editors: multi-project semantic search, CLI and MCP server.

  • Python
  • ChromaDB
  • MCP
  • Typer
  • embeddings
AI2026

Mnemosyne stores a project’s decisions, patterns and context in a local vector database so an agent or editor can retrieve them by meaning, not exact wording.

  • Local Hugging Face embeddings and ChromaDB storage.
  • Multiple projects, tags and semantic search.
  • Typer/Rich CLI and an MCP server that gives persistent memory to OpenCode, Claude Code or any MCP client.
  • Local by default: context never leaves your machine.

It is a small but very useful piece: it turns a repository’s scattered knowledge into something an agent can query across sessions.

Back to projects