Zep
Managed agent memory platform that builds temporal context graphs from conversations and business data for governed retrieval.
Quick decision
What it does
Zep is described on its homepage as a unified context layer that ingests business data, documents, and conversations, builds a Context Graph for each user, customer, account, or domain, and retrieves the context each task depends on in one policy-filtered call. The site frames this as giving agents governed context at lower token cost by returning only relevant facts instead of whole records.
The underlying open source framework Graphiti builds dynamic, temporally-aware knowledge graphs that represent evolving relationships over time, with hybrid retrieval combining time, full-text, semantic, and graph approaches. Zep Cloud uses Graphiti to derive temporal graph artifacts and runs a managed context lake, adding managed ingestion, retrieval, governance, and deployment options including bring-your-own-cloud.
The getzep/zep GitHub repository is explicitly not the product itself; it contains example code, framework integrations, ingestion tooling, ontology definitions, benchmarks, and an MCP server directory for Zep Cloud. Official SDKs are listed as zep-cloud for Python, @getzep/zep-cloud for TypeScript and JavaScript, and a Go client, and the former Community Edition is deprecated and unsupported.
Verified capabilities
Memory
Temporal context graphs Source verified
Build per-user and per-account context graphs where facts are dated and updated as source data changes, with links back to source episodes.
Ingestion
Unified ingestion Source verified
Ingest conversations, documents, and structured business data including Slack, email, JSON, and CSV through SDKs and a bulk ingestion pipeline.
Retrieval
Graph search and context assembly Source verified
Retrieve a materialized context block or raw edges, nodes, episodes, and observations ranked across scopes for direct use in LLM prompts.
Governance
Governance and policy filtering Source verified
Policies decide what context each agent can retrieve, with every request logged, plus webhooks, analytics, and custom extraction instructions on higher tiers.
Integrations
Framework integrations Source verified
Integration packages for agent frameworks including Google ADK, AutoGen, CrewAI, LangGraph, LiveKit, Pydantic AI, Mastra, and Vercel AI SDK across Python, TypeScript, and Go.
MCP
Memory MCP Server Source verified
Remote MCP endpoint that lets people use their Zep user graph from MCP clients such as Claude, ChatGPT, and Cursor with per-user sign-in.
Quick start
Install the Zep Cloud SDK
Sign up at getzep.com and install an official SDK listed in the repository README to start adding episodes and searching the context graph through the managed API.
pip install zep-cloud
# TypeScript alternative:
# npm install @getzep/zep-cloudSource: https://github.com/getzep/zep. Examples use placeholders only. Never paste a real key into a profile, config file you share, or a ticket.
MCP support: Zep documentation describes a Memory MCP Server at a remote Streamable HTTP endpoint with OAuth sign-in, also present as a zep-mcp-server directory in the repository and as seat allocations on pricing plans.
Works with
Packages
Python package zep-cloud, npm package @getzep/zep-cloud, and Go client from the getzep organization listed in the README.
Frameworks
Integration packages for Google ADK, Microsoft Agent Framework, AutoGen, CrewAI, LangGraph, LiveKit, Pydantic AI, Mastra, and Vercel AI SDK.
MCP
Remote Memory MCP Server for Claude, ChatGPT, Claude Code, Codex, and Cursor clients that support Streamable HTTP and OAuth 2.1 with PKCE.
Deployment
Managed Zep Cloud, Cloud with customer-managed encryption keys, and Bring Your Own Cloud deployment inside a customer VPC on Enterprise.
Only sourced support is listed. A missing framework means AgentsUse has not verified it yet, not that it cannot work.
Health and maintenance
4,951
Checked 2026-10-07
653
Checked 2026-10-07
Stars, forks, and Apache-2.0 license read from the fetched getzep/zep repository page, which the README states contains examples and integrations and is not the Zep product itself. The associated Graphiti framework is Apache-2.0 open source while Zep Cloud is a commercial service. PyPI version 3.28.0 from a package comparison re-read on 2026-09-08. Weekly downloads not read, so omitted.
Pricing and license
Freemium usage-based model with free trial credits each month, paid Flex and Flex Plus tiers priced per month with included credits and per-credit overage, and custom Enterprise pricing. Credits are consumed by ingestion and processing such as episode size and webhook deliveries, while memories, retrieval, and users are listed as unlimited on the pricing comparison.
Limitations and safety
- Zep Community Edition is deprecated and no longer supported, with code moved to a legacy folder, so new self-hosted use should target Graphiti or Zep Cloud instead.
- The hosted platform is commercial with credit-based pricing, and enterprise controls such as SOC 2, HIPAA BAA, audit logs, and BYOC are limited to higher tiers.
- The public GitHub repository is examples and integrations only, so product behavior must be evaluated through Zep Cloud and its documentation rather than the repo code.
Browser and data tools can read pages, fill forms, and download files. Start with a test account or read-only access, keep credentials in environment variables, and review agent actions before connecting anything that can spend money, send messages, or delete data.
Alternatives to Zep
Mem0
Open source memory layer with a simple add and search API, library and self-hosted server options, and a hosted platform.
Tradeoff: Vector and entity based retrieval with less emphasis on bi-temporal graph validity and enterprise governance than Zep context graphs.
Langfuse
Observability and evaluation platform to trace and measure agents that use external memory systems.
Tradeoff: Focuses on traces, prompts, and evals rather than providing a managed temporal memory store.
Common questions
What does Zep do for an AI agent?
Managed agent memory platform that builds temporal context graphs from conversations and business data for governed retrieval.
Is Zep open source?
This profile records the pricing model as freemium. See the pricing section for the sourced summary.
Does Zep support MCP?
Zep documentation describes a Memory MCP Server at a remote Streamable HTTP endpoint with OAuth sign-in, also present as a zep-mcp-server directory in the repository and as seat allocations on pricing plans.
Sources and freshness
- GitHub repository: https://github.com/getzep/zep
- Pricing: https://www.getzep.com/pricing/
- MCP documentation: https://help.getzep.com/use-zep-in-claude-chatgpt
- Homepage: https://www.getzep.com
Last checked 2026-10-07. Verification label: source verified, which means public claims trace to the sources above. It does not mean AgentsUse ran the tool. Spotted an error? Send a correction. Back to the tools directory.