Mem0

Mem0

Mem0 provides AI agents and assistants with a persistent, intelligent memory layer. Unlike simple conversation buffers, Mem0 uses an LLM to extract salient facts from conversations, deduplicates them, and retrieves only the most relevant memories for each new interaction — enabling truly personalized long-running agents.

Key Features

• Auto-extraction — LLM-powered fact extraction from raw conversations
• Multi-level memory — user-level, agent-level, session-level, and org-level memory scopes
• Hybrid storage — vector store (semantic search) + graph store (entity relationships) + key-value (fast facts)
• Provider-agnostic — works with OpenAI, Anthropic, Groq, and local models; Qdrant, Pinecone, Chroma for storage
• MCP server — official Mem0 MCP server exposes add_memory and search_memory as tool calls
• REST API — managed cloud option or self-hosted OSS

Quick Start

pip install mem0ai

from mem0 import Memory

mem = Memory()

# Store a memory
mem.add("I prefer Python over JavaScript", user_id="alice")

# Retrieve relevant memories
results = mem.search("What language does Alice prefer?", user_id="alice")
for r in results:
    print(r["memory"])

Install via ai-supply

npx ai-supply add mem0-agent-memory

Curated mirror of the open-source Mem0 project (Apache-2.0). Install upstream from the repository.