Mem0

Mem0 is an open-source memory layer that gives AI agents and assistants long-term memory: it extracts facts from conversations and stores them in a vector store so agents can recall user preferences and past context across sessions. Memories are scoped by user, agent, or run, with add/search/update/delete operations and hybrid retrieval (semantic + BM25 keyword + entity linking). The Python/TypeScript library runs locally with any configured LLM, while a self-hosted server and a managed cloud platform add dashboards, auth, and advanced features such as graph memory. The project is a Y Combinator S24 company with a published research paper and an open-source evaluation framework.

1. Install the library: pip install mem0ai (or npm install mem0ai).
2. Set your model keys — the defaults are OpenAI's gpt-5-mini (LLM) and text-embedding-3-small (embedder), so export OPENAI_API_KEY or reconfigure to local providers to keep memory text on your machine.
3. (Privacy) Disable the default telemetry: export MEM0_TELEMETRY=false.
4. Use it in code: from mem0 import Memory; memory = Memory(), then memory.add(messages, user_id="alice") and memory.search(query, user_id="alice") to recall before generating a response.
5. (Optional) Self-hosted server: cd server && docker compose up -d; auth is on by default — set ADMIN_API_KEY and register an admin via the wizard (or AUTH_DISABLED=true for local dev only).
6. (Optional) Managed cloud: sign up at app.mem0.ai and use the API key; the MCP endpoint is https://mcp.mem0.ai/mcp.