Back to Apps
C
ComfyUI MCP Server
by joenorton
Generate and iteratively refine AI images, audio, and video via a local ComfyUI instance.
0 stars
1 views
Works in:Cursor
Exposes:ToolsResources
What it does
Connects AI agents to a local ComfyUI instance, allowing for natural language generation and iterative refinement of visual and auditory assets. It supports a stateful control loop where agents can refer back to specific assets and modify them without re-prompting.
Tools
generate_image: Create images from a text prompt.generate_song: Create audio using tags and lyrics.regenerate: Refine an existing asset by overriding specific parameters.view_image: Provide visual feedback by injecting images into the AI's context.list_assets: Browse recently generated assets for continuity.get_job: Poll completion status of long-running generation tasks.run_workflow: Execute custom ComfyUI workflows defined in JSON.
Installation
To use with Claude Desktop, add the following to your claude_desktop_config.json:
{
"mcpServers": {
"comfyui-mcp-server": {
"command": "python3",
"args": ["/path/to/comfyui-mcp-server/server.py"]
}
}
}
Alternatively, for clients supporting HTTP/SSE, use http://127.0.0.1:9000/mcp.
Supported hosts
- claude
- cursor
Quick install
pip install -r requirements.txt && python server.pyInformation
- Pricing
- free
- Published
- 4/12/2026
