Zep

Zep is a context layer for AI agents that unifies business data, documents, and conversations into a shared, governed system. It addresses the problem of agents lacking the context they need to complete tasks correctly, by ingesting data from across a business, building a Context Graph for each user, account, or domain, and retrieving the context a task depends on in a single policy-filtered call.

The platform includes the Context Lake, which manages Context Graphs at scale, and Konig, Zep's graph database service. Graphiti handles entity extraction, relationships, and fact invalidation to construct graphs. Ingestion accepts chat, JSON, documents, and application events, while context assembly produces token-efficient context blocks for agent prompts. Retrieval is policy-filtered and provenance-linked, and governance covers attribute-based access control, multi-tenant isolation, retention policies, and audit logging. Zep integrates with agent frameworks including LangGraph, Vercel AI SDK, Pydantic AI, and Google ADK, and offers a Memory MCP Server.

Zep is aimed at teams building agents and assistants that need governed access to business context. Self-serve plans are credit-based, with a free tier offering monthly trial credits and paid Flex and Flex Plus editions; memory, retrieval, storage, and users are unmetered. Enterprise plans add custom credits, negotiated rates, a guaranteed SLA, and deployment options covering Zep Cloud, cloud with customer-managed encryption keys, or deployment in a customer VPC.