Leonardo MCP Server

🚀 Leonardo MCP Server

A Model Context Protocol (MCP) server for Leonardo AI, supporting both HTTP and stdio modes.

✨ Features

• Create image generation jobs with Leonardo AI.
• View available models.
• Check the status of image generation jobs.
• Retrieve all the user's image generation jobs.
• Support both HTTP and stdio transports.

📦 Installation

JSON Config

It supports Claude Desktop, Cursor, and other MCP clients that use JSON config files.

⚠️ Important Note

You will need to generate a Leonardo API key and set it in the environment variable LEONARDO_API_KEY before running the server.

{
  "mcpServers": {
    "leonardo-mcp-server": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/ish-joshi/leonardo-mcp-server",
        "leonardo-mcp-server",
        "stdio"
      ],
      "env": {
        "LEONARDO_API_KEY": "YOUR_LEONARDO_API_KEY"
      }
    }
  }
}

🎥 Demo

🚀 Quick Start

This server supports two running modes:

HTTP Mode (for remote clients)

• Suitability: Suitable for remote clients (e.g., ChatGPT Playground, browser-based tools).
• Endpoint: The server runs an HTTP endpoint. You must expose it to the internet if your client is remote.

Start the server in HTTP mode (default):

uvx --from git+https://github.com/ish-joshi/leonardo-mcp-server leonardo-mcp-server

Expose the local server to the internet: If your client is remote (e.g., ChatGPT Playground), you can use ngrok or a similar tunneling tool:

ngrok http 8080

Then copy the public URL from ngrok and use it as the endpoint in your client.

Stdio Mode (for local clients)

• Suitability: Suitable for local clients that communicate over standard input/output (stdio).
• Network: No network port is opened.

Start the server in stdio mode:

uvx --from git+https://github.com/ish-joshi/leonardo-mcp-server leonardo-mcp-server stdio

🔧 Technical Details

Environment Variables

🛠️ Development

• Modify Tools: Edit main.py to add or modify MCP tools.
• Run and Test: Run using python main.py and test with a compatible MCP client. I prefer to use 5ire MCP client for testing.
• More Info: See python-sdk documentation for more details.