Memory Providers
Memory providers give agents the ability to remember information across conversations and sessions. Unlike context windows that reset with each request, memory providers persist user preferences, facts, and interaction history, enabling truly personalized and continuity-aware agent experiences.
Mem0
An intelligent memory layer for AI agents and assistants that automatically extracts, stores, and retrieves relevant memories from conversations. Mem0 provides personalized interactions by remembering user preferences, facts, and context across sessions without manual configuration.
Key Features
- Automatic memory extraction — no manual tagging or storage logic required
- User, agent, and session-scoped memory isolation for multi-tenant apps
- Hybrid storage combining vector embeddings with a graph-based knowledge store
- Self-improving memory that resolves conflicts and updates stale information
Integrations
Zep
A long-term memory service for AI assistants that provides automatic conversation summarization, fact extraction, and temporal awareness. Zep enriches agent context with relevant past interactions, enabling assistants to recall and reason about prior conversations intelligently.
Key Features
- Automatic dialog summarization to compress long conversation histories
- Fact and entity extraction with temporal metadata (when things were said)
- Semantic and graph-based memory search for relevant past context
- Session management with configurable memory policies and retention
Integrations
Comparison
Comparing memory providers to help you choose the right solution for your agent's persistence needs.
| Feature | Mem0 | Zep |
|---|---|---|
| Open Source | Yes (Apache 2.0) | Yes (Apache 2.0) |
| Self-Hosting | Yes | Yes (Docker) |
| Memory Extraction | Automatic (LLM-powered) | Automatic (built-in NLP) |
| Storage Type | Vector + Graph hybrid | Vector + Structured + Graph |
| Temporal Awareness | Basic timestamps | Rich temporal metadata |
| Best For | Quick integration, automatic personalization | Complex assistants needing conversation history and facts |