Agent♥︎Age
Catalog

agentmem

Official

by oxgeneral · Python

Lightweight persistent memory for AI agents. Hybrid search. 16 tools. One SQLite file.

agentmem (io.github.oxgeneral/agentmem) MCP Server

agentmem provides lightweight persistent memory for AI agents via a Model Context Protocol (MCP) server. It stores data in a single SQLite file and supports hybrid search combining keywords and semantics. The project targets small local installs (“Zero to 12MB install”) and exposes a typed API based on 16 TypedDicts.

🛠️ Key Features

  • Persistent memory for AI agents across sessions
  • One SQLite file storage
  • Hybrid search: keywords + semantics
  • 16 tools in the MCP server
  • Typed API (16 TypedDict)

🚀 Use Cases

  • Agents that need context that survives after conversations end
  • Scenarios where each session starts from zero and memory must persist

⚡ Developer Benefits

  • Zero to 12MB install size
  • No PyTorch, no cloud, no server; local memory only
  • Production-ready: 206 unit tests and 107 quality tests on real data

⚠️ Limitations

  • Memory is provided as a single SQLite-file mechanism (“one SQLite file”).

Topics

agent-memoryai-agentfts5knowledge-graphllmmcpmcp-servermemorymodel-context-protocolpythonsqlitevector-search