@lakehouse/memory-mcp

Persistent memory for AI assistants via Model Context Protocol (MCP).

Give your AI assistant a memory that persists across conversations. Works with Claude Code, Claude Desktop, Cursor, Windsurf, and any MCP-compatible client.

Features

• Remember - Store facts, preferences, tasks, and context
• Recall - Semantic search to find relevant memories
• Forget - Remove outdated information
• Two modes:
  • Lakehouse42 (recommended) - Full semantic search, deduplication, knowledge graph
  • Local - Simple file-based storage with keyword search

Quick Start

Claude Code

Add to ~/.claude/claude_code_config.json:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["@lakehouse/memory-mcp"]
    }
  }
}

That's it! Claude Code now has persistent memory using local storage.

With Lakehouse42 Backend (Recommended)

For full semantic search capabilities, connect to Lakehouse42:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["@lakehouse/memory-mcp"],
      "env": {
        "LH42_URL": "https://api.lakehouse42.com",
        "LH42_API_KEY": "lh42_your_api_key"
      }
    }
  }
}

Get your API key at lakehouse42.com.

Claude Desktop

Add to Claude Desktop's config (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["@lakehouse/memory-mcp"],
      "env": {
        "LH42_URL": "https://api.lakehouse42.com",
        "LH42_API_KEY": "lh42_your_api_key"
      }
    }
  }
}

Cursor / Windsurf

Follow the same pattern - add the MCP server to your client's configuration.

Tools

remember

Store a memory for later recall.

Remember that the user prefers dark mode

Parameters:

• content (required) - The information to remember
• type - fact, preference, task, event, context, reflection
• importance - 0.0 to 1.0 (default: 0.5)

recall

Search memories by semantic similarity.

What are the user's preferences?

Parameters:

• query (required) - What to search for
• limit - Max results (default: 5)
• types - Filter by memory types

forget

Delete a memory by ID.

Parameters:

• memoryId (required) - ID of memory to delete
• reason - Reason for deletion

list_memories

List recent memories.

Parameters:

• limit - Max results (default: 10)

memory_status

Check memory system status and backend info.

Local vs Lakehouse42

| Feature | Local | Lakehouse42 | |---------|-------|-------------| | Persistence | ✅ JSON file | ✅ Cloud | | Search | Keyword matching | Semantic (AI-powered) | | Deduplication | ❌ | ✅ 3-tier | | Knowledge graph | ❌ | ✅ Entity relationships | | History tracking | ❌ | ✅ Full audit trail | | Multi-device sync | ❌ | ✅ |

Environment Variables

| Variable | Description | |----------|-------------| | LH42_URL | Lakehouse42 API URL (enables LH42 backend) | | LH42_API_KEY | API key for authentication | | DEBUG | Enable debug logging (true/false) |

Programmatic Usage

import { createMemoryServer, LH42Backend } from "@lakehouse/memory-mcp";

// Create server with custom config
const server = await createMemoryServer({
  lh42Url: "https://api.lakehouse42.com",
  apiKey: "lh42_xxx",
  debug: true,
});

// Or use backends directly
const backend = new LH42Backend({
  url: "https://api.lakehouse42.com",
  apiKey: "lh42_xxx",
});

await backend.initialize();
await backend.remember({ content: "User likes TypeScript" });
const results = await backend.recall({ query: "programming preferences" });

Privacy

• Local mode: All data stored in ~/.lakehouse42/memory-mcp/memories.json
• Lakehouse42 mode: Data stored securely in your Lakehouse42 account
• No data is sent to third parties
• You control your memories

License

MIT