Zep

What is Zep?

Zep is a context engineering and agent memory platform built for developers creating AI agents. It assembles relevant context from chat history, business data, and user behavior, then delivers that context to agents at retrieval speeds as low as 200ms. The platform is designed to work with any agent framework and can be integrated in as few as three lines of code. Developers at companies including Twin Health, Praktika.ai, Thrive AI Health, and Swiggy use Zep to build personalized, reliable agents without managing complex memory infrastructure themselves.

Key Features

• Agent Memory: Stores and retrieves user interactions so agents can deliver personalized responses across sessions, rather than starting fresh each time.
• Graph RAG: Uses a graph-based approach to retrieve contextually relevant information, going beyond simple vector search to understand relationships between data points.
• Agent Context: Automatically assembles context from multiple data sources, including chat history, business data, and user behavior, before passing it to the agent.
• Knowledge Graph MCP: Manages knowledge graphs to improve how agents process and connect structured information during inference.
• Open Source Option: Zep offers an open-source version for teams that want to self-host or inspect the underlying code.
• Webhooks and API Logs: Available on Flex Plus, these allow teams to monitor activity and connect Zep events to external systems (logs retained for 7 days).
• Custom Entity and Edge Types: Developers can define up to 10 custom entity and edge types on Flex, or up to 20 on Flex Plus, to tailor the knowledge graph structure to their application.

Use Cases

• Developers building AI agents: Teams use Zep to add persistent memory and context to their agents without building memory infrastructure from scratch, resulting in faster development cycles and more personalized user experiences.
• Engineering leaders implementing context-driven applications: Platform teams integrate Zep to improve response quality and throughput in production AI applications, with the goal of higher application performance and user satisfaction.

Strengths and Weaknesses

Strengths:

• Well-suited for building personalized agents quickly, according to user feedback on G2.
• Fast retrieval times are a frequently cited advantage among reviewers.
• A free tier lets developers explore the platform before committing to a paid plan.
• Unused credits roll over for up to 60 days on paid plans, reducing waste during variable workloads.

Weaknesses:

• Rate limits on the free and lower-tier plans can be restrictive for higher-volume use cases.
• The credit-based pricing model adds complexity and is harder to predict costs as usage scales.

Pricing

• Free: $0, includes 1,000 credits per month, low rate limits, and lower priority episode processing.
• Flex: $25/month, includes 20,000 credits, 600 requests per minute, 5 projects, 10 custom entity and edge types, and unlimited memories, retrieval, and users. Additional credits available at $25 per 20,000.
• Flex Plus: $475/month, includes 300,000 credits, 1,000 requests per minute, 5 projects, 20 custom entity and edge types, custom extraction instructions, webhooks, and 7-day API logs. Additional credits at $125 per 100,000.
• Enterprise: Pricing not publicly listed. Includes advanced security, compliance, and flexible deployment options. Contact Zep for details.

Unused credits roll over for up to 60 days.

FAQ

What is Zep?

Zep is a context engineering and agent memory platform built for developers creating AI agents. It assembles relevant context from chat history, business data, and user behavior, then delivers that context to agents at retrieval speeds as low as 200ms.

What is Zep used for?

Zep is used to add persistent memory and context to AI agents without building memory infrastructure from scratch. Developers use it to build personalized, reliable agents that can deliver consistent responses across sessions.

Who uses Zep?

Developers at companies including Twin Health, Praktika.ai, Thrive AI Health, and Swiggy use Zep to build AI agents. It is designed for both individual developers and engineering teams implementing context-driven applications.

What agent frameworks does Zep work with?

Zep is designed to work with any agent framework. It can be integrated in as few as three lines of code.

What are Zep's key features?

Zep includes agent memory, Graph RAG, agent context assembly, and a Knowledge Graph MCP. Developers can also define custom entity and edge types to tailor the knowledge graph structure to their application.

Does Zep offer an open-source option?

Yes, Zep offers an open-source version for teams that want to self-host or inspect the underlying code.

How fast is Zep's retrieval?

Zep delivers context to agents at retrieval speeds as low as 200ms.

What do users say about Zep's strengths?

According to user feedback on G2, Zep is well-suited for building personalized agents quickly, and fast retrieval times are a frequently cited advantage among reviewers.

What are Zep's known weaknesses?

Rate limits on the free and lower-tier plans can be restrictive for higher-volume use cases. The credit-based pricing model adds complexity and makes costs harder to predict as usage scales.

Does Zep have a free plan?

Yes, Zep offers a free tier that includes 1,000 credits per month with low rate limits and lower priority episode processing. This lets developers explore the platform before committing to a paid plan.

How much does Zep cost?

The free plan is $0 per month and includes 1,000 credits. The Flex plan starts at $25 per month and includes 20,000 credits.

Do unused Zep credits roll over?

Unused credits roll over for up to 60 days on paid plans, which reduces waste during variable workloads.

Does Zep support webhooks and API logging?

Yes, webhooks and API logs are available on Flex Plus. They allow teams to monitor activity and connect Zep events to external systems, with logs retained for 7 days.

How many custom entity and edge types can I define in Zep?

Developers can define up to 10 custom entity and edge types on the Flex plan, or up to 20 on Flex Plus.