About Zep
Zep is a context layer for AI agents. It supplies business data, user memory and skills to agents at runtime, with access rules that decide what each user and agent can read.
What it does
Zep builds a temporal knowledge graph from business data, documents and conversations. Agents query that graph for current context, and governance rules limit access per user. The platform also reviews agent runs so that skills can improve over time.
Key features
- Governed access to company knowledge
- Memory for each user
- Skills built from agent runs
- Access policies and request logs
- Customer context from many sources
Who it's for
Companies building agents for support, finance, legal or healthcare work that need controlled access to data. Developers at smaller teams may start with the open source project instead.
Pricing and good to know
The project is open source under the Apache-2.0 license. The enterprise platform is offered on request, so you start with a demo request. Retention and access settings are part of the governance model, so review them before loading sensitive data.
Highlights
- Governed access to company knowledge
- Memory for each user
- Skills built from agent runs
- Access policies and request logs
- Customer context from many sources
