What it is
Zep builds a knowledge graph from your business data, documents and conversations, then hands your AI agents the context they need to respond correctly. It tracks how facts change over time, so an agent knows what's current and what no longer applies. It's built on Graphiti, Zep's own open-source library, and delivered as a cloud service.
How to use it
- Send conversations and business data (JSON or text) to Zep through its API.
- Retrieve relevant context for each user in a single call before responding.
- Query the graph to see entities, relationships and how they changed over time.
- Plug Zep into your agent framework with SDKs for Python, TypeScript and Go.
What it's good for
- A financial assistant that must know each customer's current address or plan, not the old one.
- An insurance company whose service agent connects policies, claims and past conversations.
- A team building agents with LangGraph that needs long-term memory without designing it from scratch.
Pros
- Temporal graph separates current facts from outdated ones
- Combines conversations and structured business data in one context
- SDKs for Python, TypeScript and Go
- Its engine, Graphiti, is open source
Cons
- The full platform is cloud-only; the Community Edition was discontinued
- More complex than simple memory if you only need chat history
- Dashboard and docs are English-only
