Ship Production-Ready Agents with ZEP Memory

From Generic
To Personalized

Zep powers AI agents with agent memory built from user interactions and business data.

Get Started
Launch Playground
Book a Demo
Trusted by:

100%

Improved Agent Accuracy

Improve your agent's accuracy on complex recall tasks by up to 100%.

Read The Paper

98%

Lower Inference Costs

Decrease token usage and costs for long-term recall.

Read The Paper

90%

Lower Response Latency

Reduce your agent's response latency for long-term recall tasks.

Read The Paper

Powered By Temporal Knowledge Graphs

Build Agents That Continuously Learn & Adapt

As users interact with your agent and business, Zep integrates new knowledge, marks outdated facts as invalid, tracks changes, and retains history—enabling agents to reason about evolving user states.

See it in action
Easily Drop into LangChain

Zep's ChatHistory class makes production-ready personalization simple for any LangChain Expression Language or LangGraph app.

More Than Just Conversational Memory

Memory That Integrates With Your Business  

Send Zep a stream of chat messages and business data, and it'll intelligently fuse them into a Knowledge Graph—giving your agent accurate, relevant user facts in a single API call.

See it in action
Easily Drop into LangChain

Zep's ChatHistory class makes production-ready personalization simple for any LangChain Expression Language or LangGraph app.

Easily Define Custom Graph Ontologies

Customize Zep For Your Domain

Define domain-specific graph entities, improving the precision and relevance of agent memory. Using familiar tools such as Pydantic and Zod makes doing so simple and fast.

See it in action
Easily Drop into LangChain

Zep's ChatHistory class makes production-ready personalization simple for any LangChain Expression Language or LangGraph app.

Custom Entity

Ready for Enterprise Production and Scale

Effortlessly Scale to Millions of Users and Facts

Zep retrieves user memories instantly, even as your user count and knowledge graph grows.

SOC 2 Type 2 and Privacy Compliance

Be assured with SOC 2 Type II compliance. Zep also offers privacy controls to support CCPA and GDPR compliance.

Cloud Service or BYOC Deployment

Sign up for Zep's cloud service, or deploy on your own infrastructure to meet your security and privacy needs.

Add Agent Memory in Minutes

With minimal code changes, enable your agent with rich, relevant context from chat and business data.

Python
TypeScript
Go
 
# Add a chat message to the graph
await zep_client.memory.add(session_id, messages=[
	Message(role="Jane", role_type="user", content="I can't log in!")
])
# Add business data to the graph
await zep_client.graph.add(user_id=user_id, type="json", data={"account_status": "suspended"})

# get memory relevant to the current conversation
memory = zep_client.memory.get(session_id="session_id")
print(memory.context)

"""FACTS and ENTITIES represent relevant context to the current conversation.
# These are the most relevant facts and their valid date ranges
# format: FACT (Date range: from - to)
<FACTS>
  - Emily is experiencing issues with logging in. (2024-11-14 02:13:19+00:00 - present) 
  - Emily's account has a suspended status due to payment failure. (2024-11-14 02:03:58+00:00 - present) 
...
</FACTS>"""
 

 
// Add a chat message to the graph
await zepClient.memory.add("session_id", [
    { role: "Jane", roleType: "user", content: "I can't log in!" }
]);

// Add business data to the graph
await zepClient.graph.add(userId, "json", { account_status: "suspended" });

// Get memory relevant to the current conversation
const memory = await zepClient.memory.get("session_id");
console.log(memory.context);

```
FACTS and ENTITIES represent relevant context to the current conversation.
# These are the most relevant facts and their valid date ranges
# format: FACT (Date range: from - to)
<FACTS>
  - Emily is experiencing issues with logging in. (2024-11-14 02:13:19+00:00 - present) 
  - Emily's account has a suspended status due to payment failure. (2024-11-14 02:03:58+00:00 - present) 
...
</FACTS>
```
 

 
client.Memory.Add(context.TODO(), "sessionId", &v2.AddMemoryRequest{
	Messages: []*v2.Message{{Content: "I can't log in!", RoleType: v2.RoleTypeUser}},
})

client.Graph.Add(context.TODO(), &v2.AddDataRequest{
	UserId: userID, Type: "json", Data: map[string]interface{}{"account_status": "suspended"},
})

memory, _ := zepClient.Memory.Get(context.TODO(), "session_id")
fmt.Println(memory.Context)

```
FACTS and ENTITIES represent relevant context to the current conversation.
# These are the most relevant facts and their valid date ranges
# format: FACT (Date range: from - to)
<FACTS>
  - Emily is experiencing issues with logging in. (2024-11-14 02:13:19+00:00 - present) 
  - Emily's account has a suspended status due to payment failure. (2024-11-14 02:03:58+00:00 - present) 
...
</FACTS>
```
300ms P95 Retrieval Latency

Speed

Lightning-Fast Memory Retrieval

Zep keeps your user experience fast, recalling memories in milliseconds—perfect for building voice or text agents.

Get Started

Graphiti Graph Framework

Powered by Open-Source

Zep's open-source temporal knowledge graph library, Graphiti, serves as the core of our capability to swiftly integrate new data streams and provide comprehensive historical context regarding user states.

Open Source

Framework and Platform Agnostic

Develop agents in Python, TypeScript, or Go - with any framework or none at all.

Python

TypeScript

Go

LangChain

AutoGen

Long-Term Memory for AI Agents
Blog
Guides
DocumentationCareersTrustTermsPrivacy
Mem0 Alternative
Zep Software, Inc. © 2024. All Rights Reserved
Contact us