Zep Community Edition (legacy)

Zep Community Edition is a self-hosted memory service for developers building AI agents and conversational assistants. It turns chat history into connected facts so an assistant can recall context across conversations and distinguish current preferences from older ones.

• Memory that tracks changes: Its Graphiti knowledge graph retains historical context and records when facts become valid or stop being valid.
• Personal and shared context: Memory can belong to a session, a user, or a group, supporting both individual conversation history and shared organizational knowledge.
• Fact retrieval: Keyword, semantic, and graph search find facts relevant to the conversation, with ranking that accounts for their relationship to the user.
• Application integrations: Python and TypeScript SDKs connect memory to applications, including those built with LangChain, LangGraph, Chainlit, and Microsoft Autogen.
• Memory deletion: Applications can delete a user's memory when that context should no longer be retained.

The memory server runs on your own infrastructure, but it needs separate LLM and embedding services. It supports local servers and hosted providers with OpenAI-compatible APIs; Anthropic can connect through LiteLLM. Where conversation data goes depends on the services you connect. Community Edition doesn't include a local embedding service and uses APIs rather than a web interface. Usage telemetry is optional and can be disabled.

Zep Cloud is a separate managed service. Its dialog classification and schema-based extraction of business data from conversations are cloud capabilities.