Mem0
Open source memory layer for AI agents. Your agent forgets everything between conversations? Mem0 gives it persistent memory: user preferences, facts, conversation history, kept across sessions. Under the hood, three engines run in parallel (vector search, knowledge graph, key-value) that you never wire yourself. Works with any agent stack.
Strengths
- Three memory backends (vector, graph, key-value) orchestrated for you, no plumbing to write
- Framework-agnostic: works with LangChain, LlamaIndex, CrewAI or raw API calls
- Python and TypeScript SDKs, 21 documented official integrations
Limitations
- One more layer to host and monitor (vector store + graph) if you go self-hosted
- Overkill for a stateless chatbot: if your agent has nothing to remember, you do not need it
Best for
- Teams building an assistant or business agent that must remember a user across sessions
- Devs tired of re-injecting all the context on every call who want ready-made memory