@kengurukleo/mcp-memory-server

An MCP server that gives AI assistants persistent memory. Save and recall personal information across conversations using semantic vector search.

Works with Cursor, Claude Desktop, Codex, and any MCP-compatible client.

How It Works

When connected to your AI assistant, you can say things like:

• "Remember that my preferred editor is Cursor"
• "What do you know about my coding preferences?"
• "Forget my old email address"

Memories are stored as vector embeddings in Firebase Firestore, enabling semantic search -- you don't need exact keywords to recall them.

Setup

Prerequisites

You need a running backend (Firebase Cloud Functions). See the full setup guide for backend deployment instructions.

Once the backend is deployed, you'll have:

• A Cloud Functions URL (e.g. https://europe-west1-YOUR_PROJECT.cloudfunctions.net)
• A personal API key

Configure in Cursor

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@kengurukleo/mcp-memory-server"],
      "env": {
        "MEMORY_API_URL": "https://europe-west1-YOUR_PROJECT.cloudfunctions.net",
        "MEMORY_API_KEY": "your-api-key"
      }
    }
  }
}

Configure in Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": ["-y", "@kengurukleo/mcp-memory-server"],
      "env": {
        "MEMORY_API_URL": "https://europe-west1-YOUR_PROJECT.cloudfunctions.net",
        "MEMORY_API_KEY": "your-api-key"
      }
    }
  }
}

MCP Tools

| Tool | Description | |------|-------------| | save_memory | Save information to memory. Input: text (required), tags (optional) | | find_memories | Search memories by meaning. Input: query (required), limit (optional, 1-10) | | delete_memory | Delete a memory by ID. Input: memoryId (required) |

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | MEMORY_API_URL | Yes | Firebase Cloud Functions base URL | | MEMORY_API_KEY | Yes | Your personal API key |

Architecture

This package is a thin MCP client. It communicates via stdio with the LLM client and makes HTTPS calls to Firebase Cloud Functions which handle:

• Embedding generation (Gemini API)
• Vector storage and search (Firestore)
• API key authentication

All secrets stay server-side. This package only needs the API URL and your personal key.

License

MIT