MCP Memory

# Example Usage of the MCP Memory Tool This document demonstrates how to use the MCP Memory tool effectively in various scenarios. ## Basic Setup First, install the package with its required dependency: ```bash pip install mcp-mem hipporag ``` Then, you can run the server: ```bash mcp-mem ``` ## Example 1: Creating and Using Session Memory When an LLM needs to maintain memory across interactions: ``` Human: Let me tell you about my project. I'm building a web application using React and Node.js. LLM: I'll remember that you're building a web application using React and Node.js. [LLM uses the create_memory tool] create_memory(session_id="user123") Result: {"session_id": "user123", "status": "created", "message": "Memory for session user123 created successfully"} [LLM uses the store_memory tool] store_memory(session_id="user123", content="User is building a web application using React and Node.js", metadata={"topic": "project", "technologies": ["React", "Node.js"]}) Result: {"session_id": "user123", "memory_id": "mem123", "status": "stored", "message": "Memory stored successfully in session user123"} ``` ## Example 2: Retrieving Memory When an LLM needs to recall information from previous interactions: ``` Human: What was I working on again? [LLM uses the retrieve_memory tool] retrieve_memory(session_id="user123") Result: { "session_id": "user123", "status": "success", "query": null, "memories": [ { "id": "mem123", "content": "User is building a web application using React and Node.js", "timestamp": "2025-04-18T22:01:00", "metadata": {"topic": "project", "technologies": ["React", "Node.js"]} } ] } LLM: You were working on building a web application using React and Node.js. Would you like to continue discussing that project? ``` ## Example 3: Searching Memory When an LLM needs to find specific information: ``` Human: What technologies am I using for my project? [LLM uses the retrieve_memory tool with a query] retrieve_memory(session_id="user123", query="technologies") Result: { "session_id": "user123", "status": "success", "query": "technologies", "memories": [ { "id": "mem123", "content": "User is building a web application using React and Node.js", "timestamp": "2025-04-18T22:01:00", "metadata": {"topic": "project", "technologies": ["React", "Node.js"]} } ] } LLM: For your project, you're using React for the frontend and Node.js for the backend. ``` ## Integration with MCP Configuration To use this tool with Claude in Windsurf, add the following configuration to your MCP config file: ```json "memory": { "command": "/path/to/mcp-mem", "args": [], "type": "stdio", "pollingInterval": 30000, "startupTimeout": 30000, "restartOnFailure": true } ``` The `command` field should point to the directory where you installed the python package using pip.