MCP Memory

🚀 MCP Memory

MCP Memory is an MCP server that empowers MCP clients (including Cursor, Claude, Windsurf, etc.) to remember users' information (preferences, behaviors), which can be retained across different conversations. It leverages vector search technology to find relevant memories based on meaning rather than keywords. It is built using Cloudflare Workers, D1, Vectorize (RAG), Durable Objects, Workers AI, and Agents.

🚀 Quick Start

📺 Video

🎬 Try it Out

✨ Features

MCP Memory enables MCP clients to remember user information across different conversations and uses vector search technology to find relevant memories based on meaning.

📦 Installation

🔧 Deployment Options

Free Deployment on Cloudflare

1. Sign up for a Cloudflare account.
2. Create a new Workers application.
3. Configure Vectorize, D1 database, and Workers AI services.
4. Deploy the MCP Memory code.

Other Methods

• Run MCP Memory in a local development environment using wrangler dev.
• Ensure sufficient Cloudflare resources when deploying to a production environment.

🛠️ Setup Steps

1. Install Wrangler

   npm install -g @cloudflare/wrangler
   

2. Download the code

   git clone https://github.com/Puliczek/mcp-memory.git
   cd mcp-memory
   

3. Deploy to Cloudflare

   wrangler deploy
   

📚 Documentation

📖 How it Works

MCP Memory is implemented through the following components:

• Cloudflare Workers: Handle network requests.
• D1 Database: Store user memories.
• Vectorize: Perform vector searches to find relevant memories.
• Workers AI: Enhance search and generation capabilities.

💻 Frequently Asked Questions

1. Can I use memory.mcpgenerator.com to store my memories?

   • Yes, you can store and retrieve your memories for free through this service.
   • Your data will be securely stored and accessible only to you.
   • Note that I cannot guarantee the service will always be available.

2. Can I host MCP Memory myself?

   • Yes, you can host MCP Memory for free on Cloudflare.
   • You need a Cloudflare account and the corresponding services (Workers, Vectorize, D1 database, and Workers AI).

3. Can I run it locally?

   • Yes, you can run MCP Memory locally using wrangler dev during the development phase.
   • Note that some features such as vector search or Workers AI require a connection to Cloudflare services.

4. Can I host it elsewhere?

   • No, MCP Memory is designed for the Cloudflare architecture.

5. Why was it built?

   • I wanted an open - source solution.
   • Having control over the data is important to me.

6. Can I use it to serve multiple people?

   • Yes, MCP Memory can be integrated into your application to serve all users.
   • Each user has their own independent storage space.

7. Can I use it to store other content besides memories?

   • Yes, MCP Memory can be used to store any text - based information.
   • Practical applications include:
     • Knowledge base: Technical documentation, operating instructions, and troubleshooting guides.
     • User behavior: Track how users interact with features and common usage patterns.
     • Project notes: Record decisions and updates.

🤝 Support