Zep icon

Zep

Zep is a context engineering platform for AI agents. It builds and maintains a temporal knowledge graph from chat histories and business data, then retrieves and assembles only the most relevant context for each request. This improves accuracy and personalization while reducing hallucinations and token spend. Zep includes agent memory, Graph RAG, and context assembly services with managed cloud APIs and SDKs for Python, TypeScript/JavaScript, and Go. Developers can ingest messages and structured/unstructured records, define custom entity/edge types, and query or retrieve subgraphs on demand. Zep also ships an experimental Knowledge Graph MCP Server so MCP clients like Claude Desktop or Cursor can persist and recall context locally or against Zep Cloud. Enterprise options add SOC 2 Type II, HIPAA BAA, audit logs, SSO, SLA, and BYOC/VPC deployment.

No discussions yet

Be the first to start a discussion about Zep

Demo Video for Zep

Developer

Zep AI builds memory and context engineering infrastructure for AI agents, including a temporal knowledge graph platform (Zep), Graph R…read more

Pricing and Plans

PlanPriceFeatures
Free (Metered)Free
  • 2,500 messages free per month
  • 2.5 MB graph data free per month
  • Up to 5 projects
  • Discord/community support
Metered (Pay-as-you-go)Contact us
  • $1.25 per 1,000 messages
  • $2.50 per MB of graph data ingested
  • Unlimited retrieval requests
  • In-app chat support (with upgrade)
EnterpriseContact us
  • Custom limits & committed rate limits
  • SOC 2 Type II, HIPAA BAA
  • SSO, API & audit logs, SLA
  • Single-tenancy & BYOC/VPC deployment
  • Dedicated account manager & Slack support

System Requirements

Operating System
WINDOWS, MACOS, LINUX
Memory (RAM)
4GB+ (development)
Processor
Modern 64-bit multi-core CPU
Disk Space
100MB+ for SDKs and local caches

AI Capabilities

Long-term conversational memory
Temporal knowledge graph construction
Graph RAG over evolving data
Context retrieval and assembly
Structured data extraction from chats
Semantic and BM25 search
Entity/edge typing with custom schemas
MCP server integration for local memory