Backed By Zep
AI memory and context management platform for LLM-powered applications.
Zep is a developer tool that provides persistent memory and context management for applications powered by large language models (LLMs). It offers APIs to store, retrieve, and manage user and conversation context, enabling more personalized and context-aware AI experiences. Zep integrates with popular LLM providers and supports secure, scalable deployment for production-grade AI apps.
Persist and retrieve user context
Enhance AI app personalization
Integrate with OpenAI, Anthropic, and more
Support for multi-turn conversations
Secure and scalable deployment
API-first architecture
1,000 Episodes free per month
20,000 Credits included
Custom limits & guaranteed rate limits
Agent Memory with Temporal Knowledge Graphs (powered by Graphiti)
Graph RAG for dynamic data retrieval and business data integration
Automated Context Assembly (Context Block) for LLM prompts
Hybrid Retrieval (Vector, Keyword, Graph Traversal)
Bi-temporal Fact Tracking (Event Time and Ingestion Time)
Real user experiences from across different platforms
Zep's memory layer transformed our AI agent from a simple chatbot into a truly intelligent assistant that understands context and remembers user preferences.
Sarah Johnson
Recent
ots, support) that require precise, long-term memory.
inance) that need SOC 2/HIPAA compliance and BYOC/BYOK deployment.
ts (competitors like Graphlit may be better suited for breadth of data ingestion).
orted Community Edition (Zep OSS is now deprecated/requires more self-management).
Enables fast, accurate, and personalized AI agents with long-term memory and temporal context, significantly reducing LLM hallucinations and token costs by systematically engineering and retrieving only the most relevant, up-to-date context from a dynamic knowledge graph.
Community Edition (Zep Open Source) was deprecated in April 2025 (according to n8n docs) Conversation-first ingestion requires custom pipelines for non-chat data sources (compared to Graphlit's 30+ connectors) Open Source version requires self-management of Neo4j/FalkorDB/Kuzu and OpenSearch/Neptune infrastructure General memory retrieval from API is best suited for high-level context, and should be complemented with specialized vector/structured databases for certain needs (according to a developer review)