# ComfyUI MCP

> Drive a local ComfyUI: 40+ tools for image, video, audio and 3D workflows.

- **Category:** Model Hubs
- **Vendor:** Community (shawnrushefsky) (community)
- **Pricing:** free
- **Verified:** 2026-06-11

## What it does

Connects agents to a local or remote ComfyUI instance with 40+ tools for running, validating, composing and managing node workflows, plus 70+ bundled example workflows (FLUX, SDXL, SD3.5, ControlNet, LTX-Video, Mochi, AnimateDiff, Stable Audio). The free/local route for filmmakers who own GPUs and want full pipeline control, including open video models.

## Connect

- Claude Code: `claude mcp add comfyui -- docker run -i --rm --pull always -e COMFYUI_URL=http://host.docker.internal:8000 ghcr.io/shawnrushefsky/comfyui-mcp:latest`
- Local: `docker run -i --rm --pull always -e COMFYUI_URL=http://host.docker.internal:8000 ghcr.io/shawnrushefsky/comfyui-mcp:latest`

**Auth:** none (env `COMFYUI_URL`) — Points at your own ComfyUI instance (default port 8188) — no cloud account needed

## Capabilities

`text-to-image` · `image-to-image` · `text-to-video` · `audio-generation` · `3d-generation` · `workflow-execution` · `local-inference`

## Sample tools

`run_workflow` · `validate_workflow` · `search_templates` · `get_example_workflow` · `build_node` · `list_models` · `get_queue` · `recommend_workflow`

## Field notes

MIT; the most comprehensive ComfyUI MCP. Notable alternatives: artokun/comfyui-mcp (Claude Code plugin with slash commands/skills) and joenorton/comfyui-mcp-server (lightweight Python). Requires you to run ComfyUI yourself.

## Links

- Site: https://www.comfy.org
- Repo: https://github.com/shawnrushefsky/comfyui-mcp

---

Structured data: https://mcp.film/api/mcps/comfyui.json · Directory: https://mcp.film