Getting Started
Install the skill, connect ComfyUI, and run your first workflow.
This page is the shortest reliable path from cloning the repository to triggering a real ComfyUI workflow through OpenClaw or the local CLI.
Prerequisites
What you need before setup
- Python 3.10 or newer.
- A reachable ComfyUI server, commonly
http://127.0.0.1:8188. - At least one ComfyUI workflow exported in API format.
- A workflow output path that ends in
Save Imageor another downloadable image output.
Quick Install
Install into an OpenClaw workspace
cd ~/.openclaw/workspace/skills
git clone https://github.com/HuangYuChuh/ComfyUI_Skills_OpenClaw.git comfyui-skill-openclaw
cd comfyui-skill-openclaw
python3 -m venv .venv
source .venv/bin/activate
python3 -m pip install -r requirements.txt
cp config.example.json config.json
If your macOS Python blocks global package changes, keep everything inside the virtual environment above.
Recommended Flow
Use the local web dashboard first
- Start the UI with
./ui/run_ui.shon macOS or Linux, orui\run_ui.baton Windows. - Open
http://localhost:18189. - Add a ComfyUI server and confirm the server URL is correct.
- Upload a workflow exported from ComfyUI with Save (API Format).
- Select the parameters you want OpenClaw to expose and save the mapping.
Best practice: start with the UI once, then automate later with CLI or agent edits.
CLI Verification
Confirm the skill can see and call your workflow
First list the registered workflows:
python scripts/registry.py list
Then run a real generation test:
python scripts/comfyui_client.py \
--workflow local/Default \
--args '{"prompt":"A premium product photo"}'
Successful runs return a prompt_id and one or more local image paths in ./outputs.
Workflow Files
The minimum file layout
data/local/Default/
workflow.json
schema.json
workflow.jsonis the ComfyUI API-format export.schema.jsonexposes only the parameters the agent should be allowed to call.- The workflow identity used by the client is
<server_id>/<workflow_id>.