Comfyui Mcp
Drive ComfyUI workflows, image generation, model loading, and VRAM-aware ops from Claude Code via MCP instead of clicking the ComfyUI UI.
Overview
ComfyUI MCP is a MCP server for the Build phase that lets agents run ComfyUI workflows, generate images, and manage models and VRAM.
What is this MCP server?
- MCP server plus Claude Code plugin for ComfyUI orchestration
- Run ComfyUI workflows and generate images from agent chat
- Manage models and VRAM-related operations through tool calls
- Stdio transport via npm package comfyui-mcp (v0.4.1)
- Pairs agent-driven prompting with node-based ComfyUI pipelines
- npm package comfyui-mcp version 0.4.1
- Stdio MCP transport; ships with Claude Code plugin per server description
What problem does it solve?
Switching between Claude Code and the ComfyUI UI breaks flow when you want agent-driven image batches or workflow tweaks.
Who is it for?
Solo builders with a local or LAN ComfyUI stack who want Claude Code to operate workflows and GPU-bound generation.
Skip if: Builders without ComfyUI installed, pure cloud-only image APIs, or teams that forbid local GPU workloads.
What do I get? / Deliverables
Your agent can trigger ComfyUI graphs and generation jobs in-session so visual asset work stays scripted alongside code.
- Agent-triggered ComfyUI workflow runs and image outputs
- Model load/swap operations coordinated through MCP tools
- VRAM-aware generation sessions without manual graph edits
Recommended MCP Servers
Journey fit
Image and workflow generation happens while you are building assets, agents, and product surfaces—not during launch distribution or production monitoring. The server extends your coding agent with local ComfyUI control, which maps to build → agent-tooling rather than a one-off frontend snippet.
How it compares
ComfyUI bridge over MCP—not Midjourney-style hosted art nor a generic Stable Diffusion CLI without node graphs.
Common Questions / FAQ
Who is ComfyUI MCP for?
Developers and solo creators who already run ComfyUI and want Claude Code (or another MCP client) to run workflows and manage models programmatically.
When should I use ComfyUI MCP?
Use it while building when you need repeatable image or video workflows, asset variations, or VRAM-aware model swaps driven from your agent session.
How do I add ComfyUI MCP to my agent?
Install comfyui-mcp from npm, configure stdio MCP with your ComfyUI server URL, enable the Claude Code plugin if you use Claude, restart, then call ComfyUI tools from chat.