Zep AI

Zep - Long-term memory for AI assistants

Build Assistants that know your users.

Zep is the easiest way to add long-term memory to your AI Assistant. With Zep, you can recall, understand, and extract data from chat histories, enabling you to build rich, personalized experiences.

Sign up for the Zep Cloud and get a Zep Bot T-Shirt when you go live!


  • Zep ensures your Assistant remembers relevant facts and nuance from historical conversations, no matter how distant in the past.
  • Zep makes dialog useful: Control program flow, extract structured data, select the right prompts or tools, build customer segments, and more.
  • Fast! Minimize the number of LLM calls keeping your users waiting. Zep completes tasks in a 1/10 of the time compared to similar functionality built with OpenAI.


Zep In Action: The Amazing ShoeStoreBot

Managing sales, returns, and order inquiries is all in a day’s work for this hard-working ShoeStoreBot. It remembers past customers, their budgets, brand preferences, shoe size, and more.


Perpetual Memory:  Your Assistant will never forget a user.

Automagically populate prompts with relevant facts and summaries extracted from past conversations, no matter how distant.

Dialog Classifier: Instantly and accurately understand chat dialog.

Identify user intent, emotion, segmentation, and more. Control program flow with semantic context, using the right prompts, tools, and agents depending on the current dialog.

Extract Structured Data from Dialog

Quickly extract business data from chat conversations. Understand what your Assistant should ask for next to complete its task.

Support for Your Favorite Languages and Frameworks

Are you a Python, TypeScript, LangChain, LangChain.js, Chainlit, or FlowWise person? We got you covered.

The Zep Team

Zep was founded by Daniel (hi!), who was recently joined by 2 engineers he worked with previously (Peter & Paul!). Daniel led ML at SparkPost (acquired by MessageBird), and previously founded KnowledgeTree (another popular open source project from the 2000s). Despite taking on marketing and corp dev roles in previous lives, he remains an engineer at heart.

Daniel accidentally started Zep when he got frustrated that no delightful software existed to do what Zep does today. He built an early version and stuck it up on GitHub, and one thing led to another.