scottopolis/memory-mcp

Memory MCP Server

A Model Context Protocol (MCP) server that provides long-term memory capabilities by storing and searching text data using Qdrant vector database and OpenAI embeddings.

What It Does

This MCP server enables AI assistants to store and retrieve information semantically by:

• Storing memories: Converts text to vector embeddings using OpenAI's embedding models and stores them in Qdrant
• Semantic search: Finds similar stored content based on semantic meaning rather than exact keyword matches
• Persistent memory: Maintains information across conversations and sessions

How It Works

The server exposes two main MCP tools:

qdrant-upsert

Stores text as a memory by:

1. Converting the input text to vector embeddings using OpenAI's text-embedding-3-small model
2. Storing the vector and original text in a Qdrant collection with timestamp metadata
3. Returns a unique ID for the stored memory

qdrant-search

Retrieves similar memories by:

1. Converting the search query to vector embeddings
2. Performing cosine similarity search in Qdrant
3. Returning ranked results with similarity scores and original text

Setup

1. Install Dependencies

npm install

2. Build the Server

npm run build

3. Configure Claude Desktop

Add this to your Claude Desktop configuration file at ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "memory-mcp": {
      "command": "node",
      "args": [
        "/path/to/memory-mcp/build/index.js"
      ],
      "env": {
        "OPENAI_API_KEY": "your_openai_api_key_here",
        "QDRANT_URL": "http://localhost:6333",
        "QDRANT_API_KEY": "your_qdrant_api_key_here"
      }
    }
  }
}

Replace:

• /path/to/memory-mcp/build/index.js with the actual path to your built server
• your_openai_api_key_here with your OpenAI API key
• Qdrant URL and API key with your Qdrant configuration

4. Setup Qdrant Vector Database

You can run Qdrant locally with Docker:

docker run -p 6333:6333 qdrant/qdrant

Or use Qdrant Cloud and update the QDRANT_URL and QDRANT_API_KEY accordingly.

5. Restart Claude Desktop

After configuring, restart Claude Desktop to load the MCP server.