# Zep

> Agent context is hard. We fixed it. Zep is the context engineering platform for AI agents—with agent memory, Graph RAG, and automated context assembly. 200ms retrieval. Three lines of code.

## About Zep

Zep is a context engineering platform that enables developers to build AI agents with comprehensive memory and personalized context. The platform systematically processes user preferences, conversation history, and business data into temporal knowledge graphs that agents can use for reliable, accurate decision-making.

Context Engineering is the discipline of assembling all necessary information around an LLM to help it accomplish tasks reliably.

Unlike simple vector databases or basic memory solutions, Zep provides a complete context engineering pipeline that combines agent memory, Graph RAG, and automated context assembly to deliver enterprise-ready AI applications.

## Why Agents Fail

Agents don't fail because the model is bad. They fail because they don't have the right context.

- **Chat Memory is Chat-Only**: Blind to business data, app events, and what the user did yesterday.
- **Static RAG is Stale**: Doesn't reflect what just happened or how facts have changed.
- **Tool Calls are Unpredictable**: Slow and unreliable. The LLM decides when to call them and often guesses wrong.

Context is scattered across systems. Your agent sees pieces, not the picture. Zep fixes this.

## Value Propositions

### For Developers
Deploy personalized agent applications that understand user preferences and business context in days, not months of building retrieval infrastructure.

### For Engineering Leaders
Achieve measurable agent performance improvements (100%+ accuracy gains) with enterprise-grade personalization that combines user memory and business data while mitigating hallucination risks and compliance violations.

## Product Capabilities

### Agent Memory
Deploy agents with perfect memory that remember user preferences, past conversations, and business context across all interactions. Provides persistent, searchable memory that evolves with user relationships.
Learn more: https://www.getzep.com/product/agent-memory

### Graph RAG
Connect agents to business data with Graph RAG that understands relationships and context, automatically handling dynamic data updates. Goes beyond simple vector search to understand entity relationships and temporal changes.
Learn more: https://www.getzep.com/product/graph-rag

### Agent Context Assembly
Automated context assembly that optimizes prompts by combining relevant user memory and business context. Eliminates manual prompt engineering and ensures consistent agent performance.
Learn more: https://www.getzep.com/product/agent-context

### Open Source Memory Layer
Open source memory layer for AI applications with TypeScript and Python SDKs. Community-driven development with enterprise support options.
Learn more: https://www.getzep.com/product/open-source

## Technical Implementation

**Architecture:** Cloud-native platform with RESTful APIs
**Languages:** Python, TypeScript, Go
**Integrations:** LangChain, LangGraph, OpenAI, Anthropic, major LLM providers
**Deployment:** Hosted cloud service
**Compliance:** SOC2 Type 2, HIPAA compliant for enterprise use

## Performance Benefits

- **200ms P95 retrieval latency** for real-time voice and video agents
- **100%+ accuracy improvements** through personalized context engineering on the LongMemEval benchmark
- **90% latency reduction** with optimized context retrieval on the LongMemEval benchmark
- **98% token efficiency** by eliminating irrelevant context on the LongMemEval benchmark
- **Outperforms Mem0 by 10%** on the LOCOMO benchmark
- **Three lines of code** to production-ready agent context
- **Enterprise compliance** built-in for regulated industries

## Quick Start Example

```python
from zep_cloud.client import Zep

# Initialize Zep client
client = Zep(api_key="your-api-key")

# Add messages and retrieve context in one call
response = client.thread.add_messages(
    thread_id="thread-123",
    messages=[
        {"role": "user", "content": "I prefer vegetarian restaurants"},
        {"role": "assistant", "content": "I'll remember your vegetarian preference"}
    ],
    return_context=True
)

# Access the context block (optimized for LLM consumption)
context = response.context
```

## Common Use Cases

- **Customer Support Agents:** Remember customer history, preferences, and past issues
- **Personal AI Assistants:** Maintain long-term user relationships and preferences
- **Sales Assistants:** Track customer interactions and business relationship context
- **Enterprise Knowledge Agents:** Connect to business data with proper access controls
- **Healthcare AI:** HIPAA-compliant patient interaction memory and medical context
- **Financial Services:** Compliance-ready agents with customer relationship memory

## AI Agent Development Guides

Comprehensive guides for building effective AI agents:

- **Developer Guide To Mcp:** https://www.getzep.com/ai-agents/developer-guide-to-mcp
- **Introduction To Ai Agents:** https://www.getzep.com/ai-agents/introduction-to-ai-agents
- **Langchain Agents Langgraph:** https://www.getzep.com/ai-agents/langchain-agents-langgraph
- **Langgraph Tutorial:** https://www.getzep.com/ai-agents/langgraph-tutorial
- **Llm Evaluation Framework:** https://www.getzep.com/ai-agents/llm-evaluation-framework
- **Prompt Engineering For Reasoning Models:** https://www.getzep.com/ai-agents/prompt-engineering-for-reasoning-models
- **Reducing Llm Hallucinations:** https://www.getzep.com/ai-agents/reducing-llm-hallucinations

## Solutions by Role

**For Developers:** https://www.getzep.com/solutions/for-developers
**For Engineering Leaders:** https://www.getzep.com/solutions/for-engineering-leaders
**Context Engineering Guide:** https://www.getzep.com/solutions/context-engineering

## Resources

**Documentation:** https://help.getzep.com/
**API Reference:** https://help.getzep.com/api-reference
**GitHub Repository:** https://github.com/getzep
**Pricing:** https://www.getzep.com/pricing
**Customer Stories:** https://www.getzep.com/customers
**Security Information:** https://www.getzep.com/security
**Company Information:** https://www.getzep.com/about

## Technical Specifications

**Platform Type:** Cloud-native context engineering platform
**API Type:** RESTful APIs
**SDKs:** Python (`zep-cloud`), TypeScript (`@getzep/zep-cloud`), Go (`github.com/getzep/zep-go/v3`)
**Data Processing:** Real-time temporal knowledge graph construction
**Security:** Enterprise-grade encryption, SOC2 Type 2, HIPAA

---

**Website:** https://www.getzep.com
**Last Updated:** 2025-12-07
**Generated:** Automatically from website content