An SDK and self-hostable infrastructure for adding persistent memory to agent applications
A general-purpose memory layer with a Python/TypeScript SDK, a self-hostable server, CLI, and agent skills. It builds continuously updated memory with entity linking, BM25 + semantic search, temporal reasoning, and user/session/agent scopes. With years of maintenance behind it, it is one of the most mature infrastructures in this space.
Teams building their own agent applications or standardizing memory as a product/backend layer. For individuals who want ready-to-use coding memory, AgentMemory or Claude-Mem delivers faster results.
The new benchmark results in the README are for the managed platform; don't expect the same results from the open-source version. There are 895 open issues.
pip install mem0ai npx skills add https://github.com/mem0ai/mem0 --skill mem0