Zep AI

Zep AI provides enterprise-grade memory for AI agents using context graphs, enabling sub-200ms retrieval, governance, and provenance across a large-scale context lake. The platform emphasizes memory persistence, temporal graphs, and source-backed facts for agents used in enterprise environments.

Zep builds a Context Lake consisting of millions of context graphs governed and served as a single system. It creates memories from any source, constructs a graph, and retrieves relevant, token-efficient context. The system preserves provenance by tracing each fact back to its source episode, supports dynamic memory updates as information changes, and enables agents to reason with both current and past states. It advertises sub-200ms retrieval latency regardless of graph size, and provides policy-driven governance across all graphs and queries, including authorization, retention, and audit within the substrate.

Who it’s for: enterprise AI teams and developers building AI agents that require scalable memory, provenance, and governance for business-critical use cases

Hiring mentions and enterprise-focused features; mentions of enterprise pricing and a demo sign-up on the site