ComfyUI Skills for OpenClaw
A skill and CLI that turn saved ComfyUI workflows into agent commands, with a parameter schema and several servers
- 422 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Runs shell commands and calls your ComfyUI servers over the network, including remote ones
- Through Comfy MCP it can install nodes, download models and restart ComfyUI, but only after you agree
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/comfyui-skills-for-openclawDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Runs shell commands and calls your ComfyUI servers over the network, including remote ones
- Through Comfy MCP it can install nodes, download models and restart ComfyUI, but only after you agree
This is third-party code. Review the repository files before installing.
What it does
The skill turns saved ComfyUI workflows into commands an agent can call. The agent does not edit a raw node graph: it runs a workflow with `comfyui-skill run`, and only the parameters you described in a schema are exposed, with aliases, types and descriptions. The `comfyui-skill` CLI installs from PyPI. It imports workflow JSON (`comfyui-skill workflow import`), detects the format, builds a parameter schema, checks for missing nodes and models (`deps check`) and installs supported dependencies. Several ComfyUI servers, local and remote, share one namespace and a job goes to the right machine. Image uploads, inpainting masks, history and the job list also go through the CLI. For live discovery of templates, nodes and models, validation of a raw workflow and management of ComfyUI itself, the skill connects to the official local Comfy MCP. If the agent host already exposes its tools, the skill calls them directly, and in a shell-only host the `scripts/comfy_mcp.py` bridge covers it. An optional web UI handles setup and test runs and works through the same CLI.
Who it is for. Users of OpenClaw, Hermes Agent, Codex and Claude Code who already have ComfyUI workflows and want image generation on request in chat.
Good fit when
- You want to call a favorite ComfyUI workflow from a chat with an agent without exposing the whole node graph
- Images come in batches and jobs are queued and tracked with `comfyui-skill run` and `jobs list`
- You have a home and a remote ComfyUI server and need to know which has more VRAM: `server stats --all`
- Before running a workflow, `deps check` shows whether its nodes and models are present
Not a fit when
- You have no ComfyUI and no plan to install it: the skill calls an already running server
- You only need Comfy MCP and no workflows of your own: connecting it directly is simpler
- Your agent cannot run shell commands: the skill's main interface is the CLI
Example request
Draw a white cat with a registered ComfyUI workflow and show me the resultLimitations
- Python 3.10 or newer and a running ComfyUI server are required. Commands run from the skill folder's root: the CLI reads `config.json` and the `data/` directory from the current folder, otherwise `list` returns an empty list.
- To test a run right after import, export the workflow from ComfyUI in API format.
- Installing nodes, downloading models and restarting ComfyUI through MCP happen only after the agent explains the effect and you agree.
- The documentation marks `hermes skills install` as available once a PR is merged; for now install by cloning into the skills folder.
How to disable. Delete the skill folder from the agent's skills directory (for Claude Code `~/.claude/skills/comfyui-skill`) and, if you no longer need the CLI, run `pip uninstall comfyui-skill-cli`. Workflows and schemas live in `data/` inside the skill folder and go with it.
FAQ
How do I install ComfyUI Skills in Claude Code?
Clone the repository into `~/.claude/skills` as `comfyui-skill`, copy `config.example.json` to `config.json` and run `pip install comfyui-skill-cli`. Then `comfyui-skill server status` shows whether the CLI sees your ComfyUI server.
How do I add my own workflow?
Export it from ComfyUI in API format and run `comfyui-skill workflow import /absolute/path/workflow.json`. The CLI stores the workflow and its schema in `data/`, and `comfyui-skill run local/my-workflow --args '{"prompt": "a white cat"}'` then runs it.
How does the skill differ from the official Comfy MCP?
MCP discovers templates, nodes and models and validates raw workflows. The skill holds registered workflows, repeats runs quickly and keeps history. It calls MCP itself when live discovery is needed.
Which agents are supported?
OpenClaw, Hermes Agent, Codex, Claude Code and any agent that can run shell commands. The skill is compatible with the agentskills.io standard.
Does the skill have a web UI?
Yes, an optional one: you configure and test workflows in it before an agent uses them. It does not replace the CLI, and agent actions go through the same CLI.
Discussion
No questions or reviews yet.
Related
SkillsThe agent runs reviewed ComfyUI workflows from the command line: images, video, music, speech and object cutouts.
The official Comfy Org MCP server: an agent runs workflows in your local ComfyUI and collects the finished images
A Comfy Cloud plugin for Claude Code: commands to generate images, video, audio and 3D, connected through a hosted MCP server
A skill and MCP driver that let Claude Code, Codex and Gemini CLI run a local ComfyUI, with prompt recipes for 68 models
A lightweight Python MCP server: an agent generates and iteratively refines images, audio and video in a local ComfyUI
Skills and a local MCP server: an agent understands and creates images, video, audio, documents and 3D from a plain request.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| ComfyUI Skills for OpenClaw (this card) | no | 422 | ||
| Comfy MCP | yes | not required | 341 | Comfy Org's official MCP server that runs workflows and finds templates, with no parameter schema layer or server routing |
| ComfyUI-Agent-Kit | no | 121 | a skill and MCP driver with prompt recipes for models, with no workflow import or parameter schemas |
README in short
The page explains that the skill is a layer between the agent and ComfyUI and compares the roles of the CLI and the official Comfy MCP. Step-by-step setup covers OpenClaw, Claude Code, Codex and Hermes Agent, followed by workflow import, the web UI, multi-server management and the CLI command list. The documentation lives on GitHub Pages and is translated into five languages.
README badge
Original SKILL.md
---
name: comfyui-skill-openclaw
description: |
Run registered ComfyUI workflows through the fast comfyui-skill CLI, and use the official
local Comfy MCP for live template, node, model, validation, and orchestration capabilities.
**Use this Skill when:**
(1) The user requests to "generate an image", "draw a picture", or "execute a ComfyUI workflow".
(2) The user has specific stylistic, character, or scene requirements for image generation.
(3) The user asks you to import, register, sync, or configure saved ComfyUI workflows for later reuse.
license: Apache-2.0
metadata:
version: "1.1.0"
platforms: [macos, linux, windows]
prerequisites:
commands: ["comfyui-skill"]
env_vars: []
requires:
bins: ["comfyui-skill"]
cliHelp: "comfyui-skill --help"
hermes:
tags: [image-generation, comfyui, ai-art, workflow, stable-diffusion, flux]
related_skills: []
---
# ComfyUI Agent SKILL
> **Prerequisites**: Install the CLI: `pip install -U comfyui-skill-cli`. The official local Comfy MCP is optional and should be connected using the [official guide](https://docs.comfy.org/agent-tools/mcp#local-comfy-mcp-connection) when MCP capabilities are needed. All shell commands must run from this project's root directory (where this `SKILL.md` is located).
>
> [!IMPORTANT]
> **Directory Sensitivity**: The CLI reads `config.json` and `data/` from the current directory.
> You **MUST** `cd` into the project root before running any command.
> **Symptom**: `list` returns `[]` or `server status` reports not found → you are in the wrong directory.
## Quick Decision
- A matching registered workflow exists → use the **CLI Execution Flow (Step 1-4)**. This is the default for repeated, batch, and latency-sensitive work.
- The request has no matching workflow, or requires live template/model/node discovery → use the **official local Comfy MCP**.
- The user supplies a raw workflow that needs real environment validation → use MCP `validate_workflow`; use CLI `run --validate` for a registered workflow.
- The user asks to install nodes, download models, restart, update, or switch versions → use MCP only after explaining the effect and getting confirmation.
- User says "generate image / draw a picture" → first inspect registered workflows; if none fit, discover a current MCP template instead of guessing a model name.
- User says "import workflow / add workflow" → `comfyui-skill --json workflow import <path>`
- User says "img2img / use this image" → first `comfyui-skill --json upload <image>`, then execute
- User says "inpainting / mask this area" → `comfyui-skill --json upload <mask> --mask`, then execute
- User says "show previous results" → `comfyui-skill --json history list <id>`
- User says "what failed / check job status" → `comfyui-skill --json jobs list --status failed`
- User says "which server has more VRAM" → `comfyui-skill --json server stats --all`
- User says "what nodes are available" → use MCP `nodes` for live ecosystem discovery; use `comfyui-skill --json nodes list` when only the CLI inventory is needed.
- User says "dry run / test without executing" → `comfyui-skill --json run <id> --validate`
- User says "open management UI" → `python3 ./ui/open_ui.py`