- CollectionMemory & State
- Classificationagent-native
- MCPAvailable
- Agent SkillAvailable
- InterfacesAgent Skill · MCP · REST · SDK · API
What it does
Mem0 is a self-improving, adaptive memory layer for AI agents. It automatically extracts salient facts from conversations, reconciles new information with existing memories, and returns only the relevant context at query time — reducing token costs by 90% versus full context injection, while improving answer accuracy by 26% on memory benchmarks.
The agent calls the Mem0 API; Mem0 handles all decisions about what to store, what to update, and what to retrieve. The agent does not manage its own memory.
Agent-facing interfaces
- Catalog protocols: Agent Skill, MCP, REST, SDK, API
- MCP: ✅ Available
- Agent Skill: ✅ Official skill available via the AgentSkills ecosystem
- Onboarding type:
sdk
Protocol surface
| Interface | Detail |
|---|---|
| Python SDK | mem0.add(), mem0.search(), mem0.update(), mem0.delete() |
| REST API | Full memory CRUD and search |
| MCP Server | Model Context Protocol for direct LLM memory access |