Unraid MCP Server

A Python-based MCP (Model Context Protocol) server that enables AI assistants to interact with an Unraid server through the official Unraid GraphQL API.

Disclaimer

USE AT YOUR OWN RISK: This software provides access to your Unraid server through AI assistants. While this implementation is configured as read-only to enhance safety, you should still exercise caution when using it.

• This is an unofficial tool and is not affiliated with or endorsed by Unraid, Inc.

• All operations are limited to read-only actions to prevent system modifications

• Always maintain proper backups of your data

• Review all information and suggestions provided by AI assistants before taking action

• The developers are not responsible for any issues that may arise from using this software

Even with read-only access, monitoring tools can potentially expose sensitive system information. By using this software, you acknowledge and accept these limitations and risks.

Features

• System Information: Get detailed information about your Unraid server

• Array Management: Monitor array status

• Docker Management: List Docker containers and networks

• VM Management: List virtual machines

• Disk Information: Get detailed information about disks and unassigned devices

• Notification Management: View and manage system notifications

• Share Management: View and manage network shares

• User Management: List users

• API Key Management: List API keys

• Parity History: View parity check history

• Shares: Browse user shares on the Unraid server

• Plugins: View installed plugins and their status

• Error Handling: Comprehensive error handling with diagnostic information

• Logging: Detailed logging for troubleshooting

• Templated Resources: Access specific containers and VMs by name

Prerequisites

• Python 3.10 or later

• An Unraid server with the API enabled

• API key with appropriate permissions

Installation

Installing via Smithery

To install Unraid MCP Server for Claude Desktop automatically via Smithery:

Manual Installation

1. Clone the repository:

   git clone https://github.com/jmagar/unraid-mcp.git cd unraid-mcp

2. Create and activate a virtual environment:

   python3 -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Create a .env file with your Unraid API credentials:

   cp .env.template .env # Edit .env with your actual API URL and key

Unraid API Setup

To use this MCP server, you need to set up the Unraid API on your Unraid server:

1. Enable developer mode and the GraphQL sandbox using the CLI:

   unraid-api developer

   Follow the prompts to enable the sandbox.

2. Create an API key with the necessary permissions:

   unraid-api apikey --create

   Follow the prompts to set the name, description, roles, and permissions.

3. Configure your .env file with:

   • UNRAID_API_URL: The GraphQL URL (e.g., http://your-unraid-server-ip/graphql)

   • UNRAID_API_KEY: The API key you created

4. Test the API using the GraphQL sandbox at http://your-unraid-server-ip/graphql

Note: The Unraid API uses the x-api-key header for authentication, not Bearer tokens.

Troubleshooting

• If you get CORS errors, make sure your client includes the correct Origin header that matches the server's URL.

• Ensure your API key has the necessary roles and permissions for the queries you're trying to execute.

• Check that the GraphQL sandbox is enabled and accessible.

Usage

Running the MCP Server

Run the server in stdio mode for integration with AI assistants:

The stdio mode is useful for:

• Direct integration with AI assistants that support the MCP protocol

• Testing with the Anthropic Python SDK

• Integration with Claude in Cursor

When running in stdio mode, the server reads from standard input and writes to standard output, following the MCP protocol format. This allows for direct communication with AI assistants without requiring HTTP transport.

Server Architecture

The server is built using the FastMCP framework and consists of:

1. Unraid API Client (unraid_client.py):

   • Handles GraphQL communication with the Unraid server

   • Manages authentication and error handling

   • Provides consistent error reporting

2. MCP Server (server.py):

   • Defines resources and tools according to MCP specification

   • Exposes Unraid functionality to AI assistants

   • Handles request validation and error diagnostics

Available Resources

Available Tools

System Management

Array Management

Docker Management

VM Management

Notification Management

Share Management

Disk Management

User Management

API Key Management

Integration with Claude

To use the MCP server with Claude API or other AI assistants that support stdio mode:

1. Create a configuration file (e.g., unraid_mcp_config.json):

   { "mcpServers": { "unraid": { "command": "/path/to/python", "args": ["/path/to/unraid-mcp/run_server.py"], "env": { "UNRAID_API_URL": "http://your-unraid-server:port/graphql", "UNRAID_API_KEY": "your-api-key", "LOG_LEVEL": "INFO", "CLAUDE_MCP_SERVER": "true" }, "disabled": false, "autoApprove": [] } } }

Note: For Windows users, make sure to use double backslashes in paths (e.g., C:\\Users\\username\\unraid-mcp\\run_server.py)

Example Queries

• "What's the current CPU usage on my Unraid server?"

• "List all of my Docker containers"

• "Tell me about my Plex container" (uses container_details resource)

• "Start the Plex container"

• "What's the status of my array?"

• "How much free space do I have on my Unraid server?"

• "Show me details about my Windows VM" (uses vm_details resource)

• "What plugins do I have installed?"

Troubleshooting

Check the log file (unraid_mcp.log) for detailed error information.

Common issues:

• Incorrect API URL or key in .env file

• Network connectivity issues to Unraid server

• Insufficient permissions for the API key

• Developer mode not enabled on Unraid server

• API key not having the necessary roles

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

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

References

• Model Context Protocol

• MCP Python SDK

• Unraid API Documentation