Zep
Long-term memory service for AI agents and LLM applications. Zep stores, enriches, and retrieves conversation history and user facts for agents. Automatically extracts facts, preferences, and summaries from conversations, building a per-user memory graph. Enables agents to remember users across sessions and retrieve contextually relevant memories at inference time.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
SOC2 certified for Zep Cloud. User conversation data stored — data handling agreements important for production. Self-hosted keeps data in your infrastructure. HTTPS enforced for Cloud.
⚡ Reliability
Best When
You're building multi-session AI assistants or agents that need to remember individual users across conversations, extract facts, and personalize responses over time.
Avoid When
Your agent is stateless or doesn't need per-user memory — conversation context can be passed directly in the prompt for simpler use cases.
Use Cases
- • Give AI agents persistent memory across conversation sessions so they remember user preferences, past interactions, and stated facts
- • Extract structured facts from agent conversations (user location, preferences, past decisions) for automatic personalization
- • Retrieve contextually relevant memories at the start of each agent session to provide personalized context injection
- • Build user knowledge graphs from agent interactions — Zep automatically links related facts and entities
- • Replace manual conversation summarization with automatic Zep summaries that preserve important information as context grows
Not For
- • Document RAG pipelines over external knowledge bases — use LlamaIndex, LangChain, or Haystack for document retrieval; Zep is for agent memory
- • Short-lived stateless agent tasks without user context — Zep adds value for multi-session, personalized agent workflows
- • Teams that need on-premises only deployment — Zep Cloud is available; self-hosted requires significant infrastructure
Interface
Authentication
Zep Cloud uses API key (passed as Z-API-Key header). Self-hosted Zep uses configurable auth. SDK handles auth automatically when API key is provided at initialization.
Pricing
Zep Cloud has a generous free tier for development. Self-hosted Zep is open source and free. Pricing is based on stored memory facts, not API calls — predictable for long-running agents.
Agent Metadata
Known Gotchas
- ⚠ Memory extraction is asynchronous — after adding messages, facts are not immediately available; allow 1-5 seconds before querying extracted facts
- ⚠ Zep v2 has significant API changes from v1 — ensure SDK version matches your Zep server version
- ⚠ Session IDs and User IDs must be consistent across requests — agents must manage these identifiers carefully to maintain memory continuity
- ⚠ Fact extraction quality depends on conversation quality — vague or ambiguous agent/user exchanges may not extract useful structured facts
- ⚠ Memory search uses semantic similarity — queries must be natural language questions, not keyword filters
- ⚠ Self-hosted Zep requires PostgreSQL with pgvector extension — infrastructure setup is non-trivial
Alternatives
Full Evaluation Report
Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for Zep.
AI-powered analysis · PDF + markdown · Delivered within 30 minutes
Package Brief
Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.
Delivered within 10 minutes
Score Monitoring
Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.
Continuous monitoring
Scores are editorial opinions as of 2026-03-07.