artokun/comfyui-mcp

Local-first, agent-native control plane for ComfyUI — MCP server + sidebar agent that generates images, video & audio, authors and runs workflows, and edits your live graph in natural language on ANY LLM (Claude, ChatGPT, Gemini, offline Ollama, or any hosted model). 178 tools, 36 AI skills, 55 installer packs. Local, LAN, VPS, or Comfy Cloud.

What it solves

comfyui-mcp provides a comprehensive control plane for ComfyUI, allowing users to drive the node-based generative AI interface using natural language via any LLM. It moves beyond simple prompt-forwarding by enabling an AI agent to actually operate the ComfyUI environment—authoring, editing, and managing workflows, models, and custom nodes—rather than just triggering a render.

How it works

The project implements a Model Context Protocol (MCP) server that acts as a bridge between an LLM (such as Claude, ChatGPT, Gemini, or local models via Ollama) and a ComfyUI installation. It can target local installs, LAN, VPS, or Comfy Cloud.

Key components include:

  • MCP Server: A standalone server that exposes 38 tools to the LLM for workflow execution and graph manipulation.
  • Agent Panel: A sidebar extension for ComfyUI that allows an autonomous agent to drive the live canvas in real-time.
  • Claude Code Plugin: Adds specialized slash commands, autonomous agents (explorer, debugger, optimizer, researcher), and hooks (like a VRAM watchdog) to the Claude Code environment.
  • Skill Packs: Bundled expertise for specific model families (e.g., Flux, WAN, LTX 2.3) that provide the agent with the correct samplers, CFG, and resolutions without trial and error.

Who it’s for

Generative AI artists and developers who use ComfyUI and want to automate complex workflow creation, diagnose failures, or manage their environment using an AI assistant across various deployment shapes (local GPU, remote pods, or cloud).

Highlights

  • Full Graph Control: Edits the live ComfyUI graph node-by-node in natural language.
  • Multi-LLM Support: Compatible with Claude, ChatGPT, Gemini, Ollama, and any OpenAI-compatible endpoint.
  • Autonomous Agents: Includes specialized agents for debugging workflow failures and optimizing VRAM usage.
  • Model Expertise: Ships with curated guides and installer packs for 13+ model families to ensure high-quality generation.
  • Remote Orchestration: Ability to drive a remote ComfyUI instance from a local machine via a secure tunnel.
  • VRAM Watchdog: A pre-execution hook that warns the user if GPU memory is insufficient for the requested task.

Related

  • Project
    CoplayDev/unity-mcp

    A bridge that connects AI assistants to the Unity Editor via the Model Context Protocol, enabling developers to manage scenes, assets, and scripts using natural language.

  • Project
    heshengtao/comfyui_LLM_party

    A comprehensive set of ComfyUI nodes for building LLM workflows, enabling the integration of AI agents, RAG, and VLMs into visual image generation pipelines.

  • Project
    IvanMurzak/Unity-MCP

    An AI-powered game development assistant that connects AI agents to Unity via the Model Context Protocol to automate editor workflows and enable in-game AI behavior.

  • Project
    ATH-MaaS/ComfyUI-Copilot

    An intelligent assistant for ComfyUI that automates the generation, debugging, and optimization of image generation workflows using an LLM-powered agent.

  • Project
    CursorTouch/Windows-MCP

    An MCP server that enables AI agents to natively control the Windows OS, perform UI automation, and manage files without needing computer vision.