ComfyUI MCP Server

Overview

The ComfyUI MCP Server exposes ComfyUI image generation capabilities to your AI assistants via the Model Context Protocol (MCP). This allows LLMs like Claude, Cursor, or any MCP-compatible client to generate images through natural language requests.

Forked from the original by Karim Lalani with enhancements for production use.

Features

• Natural language image generation - Ask your AI assistant to create images
• Docker-ready - Pre-built images on Docker Hub
• Prompt enhancement - Optional Ollama integration for better prompts
• Context updates - Periodic status updates prevent timeouts on long generations
• Seed control - Randomize or specify seeds for reproducible results
• CVE-hardened - All critical and high-priority vulnerabilities addressed

Quick Start

Docker (Recommended)

The fastest way to get started:

docker run -d \
  -e COMFY_URL=http://your-comfy-server:8188 \
  -e COMFY_WORKFLOW_JSON_FILE=/workflow.json \
  -e PROMPT_NODE_ID=6 \
  -e OUTPUT_NODE_ID=9 \
  -e SEED_NODE_ID=9 \
  -v /path/to/workflow.json:/workflow.json \
  snowgoons/comfyui-mcp-server:latest

Docker Hub →

Configuration

Optional Ollama Integration:

MCP Client Configuration

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "ComfyUI MCP": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e", "COMFY_URL=http://your-comfy-server:8188",
        "-e", "COMFY_WORKFLOW_JSON_FILE=/workflow.json",
        "-e", "PROMPT_NODE_ID=6",
        "-e", "OUTPUT_NODE_ID=9",
        "-e", "SEED_NODE_ID=9",
        "-v", "/path/to/workflow.json:/workflow.json",
        "snowgoons/comfyui-mcp-server:latest"
      ]
    }
  }
}

Cursor IDE

Add to your .cursor/mcp.json:

{
  "mcpServers": {
    "ComfyUI MCP": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e", "COMFY_URL=http://your-comfy-server:8188",
        "-e", "COMFY_WORKFLOW_JSON_FILE=/workflow.json",
        "-e", "PROMPT_NODE_ID=6",
        "-e", "OUTPUT_NODE_ID=9",
        "-e", "SEED_NODE_ID=9",
        "-v", "/path/to/workflow.json:/workflow.json",
        "snowgoons/comfyui-mcp-server:latest"
      ]
    }
  }
}

Usage

Once configured, your AI assistant can generate images with simple requests:

“Create an image of a futuristic cityscape at sunset”

The server will:

1. Submit the prompt to your ComfyUI instance
2. Monitor generation progress
3. Return the completed image

Functions Available

generate_image(prompt: str, seed: int | null)

• prompt: Text description of desired image
• seed: Optional seed value for reproducibility (null = random)

generate_prompt(topic: str) (requires Ollama)

• Enhances simple topics into detailed image prompts

Getting Your Workflow JSON

1. Build your workflow in ComfyUI
2. Click the API Export button
3. Save the JSON file
4. Mount it into the container or specify its path

Version History

v0.2.2

• Addressed all Critical and High priority CVE dependencies

v0.2.1

• Release process update

v0.2.0

• Added periodic MCP ContextInfo updates to prevent client timeouts
• Added COMFY_TIMEOUT environment variable
• Fixed caching issue with repeated prompts/seeds
• Added seed randomization support

Resources

• GitLab Repository
• Docker Hub
• Original Project
• Model Context Protocol

License

MIT License - see the LICENSE file for details.

Article by Olly ([email protected]) - Generated with local AI