This project provides a local-first, agent-native control plane for ComfyUI, enabling users to drive ComfyUI workflows using any LLM, whether cloud-based or local. It offers an extensive suite of tools and AI skills, allowing for natural language control over image/video generation, workflow authoring, model management, and live graph editing, significantly enhancing ComfyUI's automation capabilities.

How It Works

The system operates as an MCP (Model Context Protocol) server that integrates deeply with ComfyUI's REST API and WebSocket interfaces. It exposes over 100 tools and 30 AI skills that LLMs can invoke to perform complex tasks. This architecture allows LLMs to dynamically build, execute, and modify ComfyUI workflows, manage models, and troubleshoot issues, abstracting the node-based complexity through natural language commands. It supports a wide range of LLM providers and deployment configurations, from local setups to cloud instances.

Quick Start & Requirements

• Primary install/run command: npx -y comfyui-mcp (no repository clone needed for basic usage). Can also be installed as a Claude Code plugin.
• Prerequisites:
  • ComfyUI must be installed and running.
  • Node.js >= 22.0.0 (for development).
  • An LLM with tool-calling capabilities (e.g., Claude, ChatGPT, Gemini, Ollama).
  • Optional environment variables for enhanced functionality: CIVITAI_API_TOKEN, HUGGINGFACE_TOKEN, GITHUB_TOKEN.
• Links:
  • Full documentation: comfyui-mcp.artokun.io/docs
  • Claude Code plugin: Search "comfyui-agent-panel" in ComfyUI-Manager or use /plugin install artokun/comfyui-mcp.

Highlighted Details

• Extensive Tooling: Offers 108 MCP tools for workflow execution, generation, model management, diagnostics, and process control.
• AI Skills & Agents: Features 32 AI skills (e.g., model-specific guides, troubleshooting) and 4 autonomous agents (explorer, debugger, optimizer) to enhance LLM-driven ComfyUI operations.
• Multi-LLM & Deployment: Supports diverse LLMs (local via Ollama, cloud via APIs) and deployment modes (local, LAN, VPS, Comfy Cloud) from a single configuration.
• Agent Panel: Integrates an "Agent Panel" into ComfyUI's sidebar for direct LLM interaction and live graph control.
• Advanced Workflow Management: Includes visualization (Mermaid), format conversion (UI/API), comparison, debugging, and extraction of workflows from generated images.

Maintenance & Community

• Maintainer: @artokun, a known contributor within the Comfy-Org ecosystem.
• Community: Specific Discord/Slack links are not provided in the README.
• Roadmap: Mentions planned features like OAuth but no explicit roadmap document is linked.

Licensing & Compatibility

• License: MIT License.
• Compatibility: Permissive MIT license generally allows for commercial use and integration with closed-source projects.

Limitations & Caveats

• Some advanced features are provider-specific (e.g., conversation rollback for Claude).
• Local LLMs require robust tool-calling support; older or incompatible models may not function.
• Comfy Cloud integration is currently in an invite-only beta phase.
• The Agent Panel requires enabling an "external/local orchestrator" setting within ComfyUI.