ComfyUI MCP Server

A server implementation for integrating ComfyUI with MCP.

⚠️ IMPORTANT: Required ComfyUI Server This MCP server requires an active ComfyUI server to function. You must either:

1. Host your own ComfyUI server
2. Have access to an existing ComfyUI server address

Without a running ComfyUI server, this MCP server will not work properly.

Debugging

ComfyUI Debugging

python src/test_comfyui.py

MCP Debugging

mcp dev src/server.py

Installation and Configuration

ComfyUI Configuration

To modify ComfyUI host and port settings, edit src/.env:

COMFYUI_HOST=localhost
COMFYUI_PORT=8188

Custom Workflows

To add new tools, place your workflow JSON files in the workflows directory. Each new workflow must be declared as a new tool in the system.

Built-in Tools

text_to_image

The built-in text_to_image tool returns only the URL of the generated image. To obtain the actual image, you can:

• Use the download_image tool to save the image
• Access the URL directly in your browser

download_image

You can use this tool to download images generated by other tools like text_to_image. Simply provide the image URL returned from the previous tool, and it will handle the download process for you.

1. Using UV (Recommended)

Configure mcp.json:

{
  "mcpServers": {
    "comfyui": {
      "command": "uv",
      "args": [
        "--directory",
        "PATH/MCP/comfyui",
        "run",
        "--with",
        "mcp",
        "--with",
        "websocket-client",
        "--with",
        "python-dotenv",
        "mcp",
        "run",
        "src/server.py:mcp"
      ]
    }
  }
}

2. Using Docker

Build Docker Image

docker image build -t mcp/comfyui .

MCP Configuration

Configure mcp.json:

{
  "mcpServers": {
    "comfyui": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-p",
        "3001:3000",
        "mcp/comfyui"
      ]
    }
  }
}