explaingit

comfy-org/comfy-skills

13Audience · vibe coderComplexity · 2/5Setup · easy

TLDR

comfy-skills is the official Claude Code plugin that connects AI coding agents to Comfy Cloud, adding slash commands for generating images, video, audio, 3D models, background removal, and upscaling via a hosted MCP server.

Mindmap

mindmap
  root((comfy-skills))
    What it does
      Adds generation commands
      Connects to Comfy Cloud
      Uses MCP server protocol
    Commands
      Generate images
      Generate video
      Generate audio
      Remove backgrounds
    Setup
      Marketplace install
      OAuth sign-in
      Two commands total
    Audience
      Vibe coders
      Designers
      AI users
Click or tap to explore — scroll the page freely

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

Things people build with this

USE CASE 1

Add AI image, video, audio, and 3D model generation directly to Claude Code sessions with just two install commands.

USE CASE 2

Generate images from text prompts inside your AI coding workflow without switching to a separate tool.

USE CASE 3

Remove backgrounds from images or upscale them by issuing a slash command during a Claude Code chat.

USE CASE 4

Connect other MCP-compatible AI clients like Claude Desktop to the same Comfy Cloud generation tools.

Tech stack

MCPClaude CodeOAuthComfyUI

Getting it running

Difficulty · easy Time to first run · 5min

Requires a Comfy Cloud account to authenticate via OAuth after running the two install commands.

In plain English

This repository is the official home for Comfy Cloud's plugin and slash commands that connect AI coding agents, primarily Claude Code, to the Comfy Cloud image and media generation platform. ComfyUI is a popular tool for creating images and other media using AI image models, and Comfy Cloud is the hosted version of that system. This repository packages the connection between the two into a single installable plugin. Installing the plugin in Claude Code takes two commands: one to add it from the marketplace and one to sign in with OAuth. After that, slash commands become available for generating images, generating video, generating audio, generating 3D models, removing backgrounds from images, upscaling images, searching available models and processing nodes, searching workflow templates, and combining photos. A help command lists everything the tools can do. The technical mechanism is a hosted MCP server, which is a standardized protocol that lets an AI agent call external tools. The plugin bundles the server address and the command definitions together so the agent does not need to be manually configured. Other AI clients like Claude Desktop can also connect to the same hosted MCP server through their own configuration, with instructions documented on the Comfy website. The repository includes guidance for anyone contributing new commands. The key rule is that commands should describe an approach rather than listing specific model names, template names, or node identifiers directly. Because a command is frozen at the time a user installs it, hardcoding specific names would cause the commands to become inaccurate as the catalog changes. Instead, commands should tell the agent to use the search tools at runtime to discover what is currently available. The folder structure separates the active Claude Code plugin in the claude-code/ directory from a legacy flat layout in skills/ that is kept only for an older deprecated installer and will eventually be removed.

Copy-paste prompts

Prompt 1
I installed comfy-skills in Claude Code. Generate a product photo of a wooden chair on a white background, then remove the background and upscale the result to 4K.
Prompt 2
Using comfy-skills, search for available image generation models, then generate a short looping video of a sunset over the ocean using the best available model.
Prompt 3
I want to combine two portrait photos using comfy-skills. Which slash command handles photo combining, what inputs does it expect, and can you show me an example?
Open on GitHub → Explain another repo

← comfy-org on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.