MinIO Model-Context Protocol (MCP)

This project implements a Model-Context Protocol (MCP) server and client for MinIO object storage. It provides a standardized way to interact with MinIO.

Features

Server

Resources

Exposes MinIO data through Resources. The server can access and provide:

• Text files (automatically detected based on file extension)

• Binary files (handled as application/octet-stream)

• Bucket contents (up to 1000 objects per bucket)

Tools

• ListBuckets

  • Returns a list of all buckets owned by the authenticated sender of the request

  • Optional parameters: start_after (pagination), max_buckets (limit results)

• ListObjects

  • Returns some or all (up to 1,000) of the objects in a bucket with each request

  • Required parameter: bucket_name

  • Optional parameters: prefix (filter by prefix), max_keys (limit results)

• GetObject

  • Retrieves an object from MinIO

  • Required parameters: bucket_name, object_name

• PutObject

  • Uploads a file to MinIO bucket using fput method

  • Required parameters: bucket_name, object_name, file_path

Client

The project includes multiple client implementations:

1. Basic Client - Simple client for direct interaction with the MinIO MCP server

2. Anthropic Client - Integration with Anthropic's Claude models for AI-powered interactions with MinIO

Related MCP server: MCP File Server

Installation

1. Clone the repository:

git clone https://github.com/yourusername/minio-mcp.git
cd minio-mcp

2. Install dependencies using pip:

pip install -r requirements.txt

Or using uv:

uv pip install -r requirements.txt

Environment Configuration

Create a .env file in the root directory with the following configuration:

# MinIO Configuration
MINIO_ENDPOINT=play.min.io
MINIO_ACCESS_KEY=your_access_key
MINIO_SECRET_KEY=your_secret_key
MINIO_SECURE=true
MINIO_MAX_BUCKETS=5

# Server Configuration
SERVER_HOST=0.0.0.0
SERVER_PORT=8000

# For Anthropic Client (if using)
ANTHROPIC_API_KEY=your_anthropic_api_key

Usage

Running the Server

The server can be run directly:

python src/minio_mcp_server/server.py

Using the Basic Client

from src.client import main
import asyncio

asyncio.run(main())

Using the Anthropic Client

1. Configure the servers in src/client/servers_config.json:

{
  "mcpServers": {
    "minio_service": {
      "command": "python",
      "args": ["path/to/minio_mcp_server/server.py"]
    }
  }
}

2. Run the client:

python src/client/mcp_anthropic_client.py

3. Interact with the assistant:

   • The assistant will automatically detect available tools

   • You can ask questions about your MinIO data

   • The assistant will use the appropriate tools to retrieve information

4. Exit the session:

   • Type quit or exit to end the session

Integration with Claude Desktop

You can integrate this MCP server with Claude Desktop:

Configuration

On MacOS: ~/Library/Application\ Support/Claude/claude_desktop_config.json
On Windows: %APPDATA%/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "minio-mcp": {
      "command": "python",
      "args": [
        "path/to/minio-mcp/src/minio_mcp_server/server.py"
      ]
    }
  }
}

Development

Project Structure

minio-mcp/
├── src/
│   ├── client/                  # Client implementations
│   │   ├── mcp_anthropic_client.py  # Anthropic integration
│   │   └── servers_config.json  # Server configuration
│   ├── minio_mcp_server/        # MCP server implementation
│   │   ├── resources/           # Resource implementations
│   │   │   └── minio_resource.py  # MinIO resource
│   │   └── server.py            # Main server implementation
│   ├── __init__.py
│   └── client.py                # Basic client implementation
├── LICENSE
├── pyproject.toml
├── README.md
└── requirements.txt

Running Tests

pytest

Code Formatting

black src/
isort src/
flake8 src/

Debugging

Since MCP servers run over stdio, debugging can be challenging. For the best debugging experience, we recommend using the MCP Inspector:

npx @modelcontextprotocol/inspector python path/to/minio-mcp/src/minio_mcp_server/server.py

Upon launching, the Inspector will display a URL that you can access in your browser to begin debugging.

License

This project is licensed under the MIT License - see the LICENSE file for details.