yynps737/mcp-memory-server

MCP Memory Server

A Model Context Protocol (MCP) compliant memory management server for Claude Code.

Features

• ✅ Full MCP specification compliance
• ✅ Intelligent memory management with mem0
• ✅ Vector storage with Qdrant
• ✅ Optional Redis caching
• ✅ OpenAI embeddings (text-embedding-3-large)
• ✅ Semantic search capabilities

Quick Start

1. Setup environment:

   cp .env.example .env
   # Edit .env with your OPENAI_API_KEY
   

2. Start the server:

   ./start.sh
   

3. Stop the server:

   ./stop.sh
   

Project Structure

mem0-service/
├── mcp_memory_server.py  # Main MCP server
├── mcp.json             # MCP configuration
├── requirements.txt     # Python dependencies
├── start.sh            # Start script (uses uv)
├── stop.sh             # Stop script
├── .env.example        # Environment template
├── .gitignore          # Git ignore rules
├── README.md           # This file
├── data/               # Data storage (auto-created)
├── logs/               # Log files (auto-created)
└── .venv/              # Virtual environment (uv)

MCP Tools

The server provides 7 memory management tools:

1. add_memory - Add new memories
2. search_memories - Semantic search with caching
3. get_all_memories - Retrieve all user memories
4. update_memory - Update existing memories
5. delete_memory - Delete memories
6. get_memory_history - View memory change history
7. store_raw_memory - Direct storage without LLM processing

MCP Resources

• memory://stats - System statistics
• memory://config - Current configuration

Configuration

Configure Claude Desktop to use this MCP server by adding to your MCP settings:

{
  "mcpServers": {
    "memory": {
      "command": "python",
      "args": ["mcp_memory_server.py"],
      "cwd": "/path/to/mem0-service"
    }
  }
}

Dependencies

• Python 3.10+
• uv (for virtual environment management)
• Qdrant v1.14.1 (included)
• Redis (optional, for caching)

License

MIT