comfyui-mcp-server

Overview

ComfyUI MCP Server provides a bridge between AI agents and local ComfyUI installations, enabling natural language control over media generation. The server exposes various tools including image generation, asset management, job control, and workflow execution. It supports iterative refinement through its regenerate functionality and provides asset identity for reliable follow-up operations. The architecture uses a streamable HTTP transport and supports both simple single-shot generation and complex stateful control loops for more advanced use cases.

Try asking AI

After installing, here are 8 things you can ask your AI assistant:

• generate_image

  Generate images from text prompts

  Try:Generate a beautiful sunset over mountainsTry:Create a cat sitting on a mat
• generate_song

  Generate audio from tags and lyrics

  Try:Generate a song with jazz tags and lyrics about loveTry:Create an electronic track with techno tags
• regenerate

  Regenerate an existing asset with parameter overrides

  Try:Regenerate the image with brighter colorsTry:Recreate the song with a different tempo
• view_image

  View generated images inline in the conversation

  Try:Show me the generated imageTry:Display the latest artwork
• get_job

  Poll job completion status by prompt_id

  Try:Check if my image generation is completeTry:Get status of the song creation job
• list_assets

  Browse recently generated assets

  Try:Show me all generated imagesTry:List recent audio files
• get_asset_metadata

  Get provenance and parameters for an asset

  Try:Get details about this imageTry:Show metadata for the song
• cancel_job

  Cancel a queued or running job

  Try:Cancel the current image generationTry:Stop the audio creation process
• list_models

  List available ComfyUI models

  Try:Show available image generation modelsTry:List all music generation models
• set_defaults

  Set default values for generation parameters

  Try:Set default image size to 512x512Try:Make 20 steps the default for generations
• list_workflows

  List all available ComfyUI workflows

  Try:Show available workflowsTry:List all image generation workflows
• publish_asset

  Publish a generated asset to the web directory

  Try:Publish this image as hero.webpTry:Share the song on the website

Comparable tools

Installation

git clone https://github.com/joenorton/comfyui-mcp-server.git
cd comfyui-mcp-server
pip install -r requirements.txt

cd <ComfyUI_dir>
python main.py --port 8188

python server.py

{
  "mcpServers": {
    "comfyui-mcp-server": {
      "type": "streamable-http",
      "url": "http://127.0.0.1:9000/mcp"
    }
  }
}

FAQ

What is the difference between generate_image and regenerate?

generate_image creates a new image from a prompt, while regenerate recreates an existing asset with optional parameter overrides without requiring a new prompt.

Can I use custom ComfyUI workflows with this server?

Yes, place JSON workflow files in the workflows/ directory with PARAM_* placeholders to expose parameters as MCP tools.

How are assets identified and referenced?

Assets are identified by (filename, subfolder, type) tuples instead of URLs for robustness across hostname changes.

Compare comfyui-mcp-server with