Guide · Agent skill

How to use the ComfyUI Command-Line Control agent skill

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

The cli-anything-comfyui skill teaches an AI agent to run ComfyUI, an AI image generation tool, through a command-line interface instead of its graphical window. This guide explains what the skill does, how to install it as an agent skill, how to ask for it in plain language, and what to check for safety. By the end you will be able to confirm your ComfyUI server is reachable and have an agent queue a workflow and fetch the finished images.

ComfyUI Command-Line ControlHKUDS/CLI-Anything · Apache-2.0 Download the skill

Quick start

1Install ComfyUI and start it. Make sure ComfyUI is installed and running at http://localhost:8188 before using the skill.
2Install Python 3.10 or newer. The CLI requires Python 3.10+. Check your version before installing the package.
3Install the cli-anything-comfyui package. Run pip install cli-anything-comfyui to install the command-line tool.
4Add the skill folder. Unzip the downloaded ZIP into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
5Verify the server. Run cli-anything-comfyui system stats to confirm ComfyUI responds.
6Ask for work in plain language. Request a task such as listing checkpoints or queuing a workflow, and the agent will run the matching commands.

What an agent skill is and what this one does

An agent skill is a folder containing a SKILL.md file that describes a task and how to do it. When installed, an AI agent such as Claude can read the file and use the commands it describes when a request matches.

This skill covers ComfyUI through its REST API. Once installed, the agent can check the server, list available models, load and validate workflow JSON files, queue generations, monitor the queue, and download generated images. It is written for agents and power users who want to automate image generation without a GUI.

Install the skill and the CLI

Download the skill ZIP from this page and unzip it into one of these locations: ~/.claude/skills/ for your user account, or a project's .claude/skills/ folder for one project. You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills.

The skill uses a separate command-line package. Install it with pip, then confirm it works.

Python 3.10 or newer is required.
ComfyUI must be installed and running at http://localhost:8188 before the skill can do anything.
The skill package is a single SKILL.md file, with the README.md, TEST.md and HARNESS.md documents referenced as living in the cli-anything-comfyui package.

pip install cli-anything-comfyui

Check that ComfyUI is running

Before any other command, confirm the server responds. The source advises verifying ComfyUI is running with the system stats command. If this fails, start ComfyUI first; the skill cannot work without it.

You can also view the command help and start the interactive REPL mode, which supports tab-completion and history.

Show help with --help.
Start the REPL with the repl command for exploratory use.

cli-anything-comfyui system stats

Commands the agent can use

Commands are grouped into workflow, queue, models, images and system areas. Workflow commands list saved workflows, load a workflow from a JSON file, and validate it against the ComfyUI node graph. Queue commands submit a workflow, show running and pending items, clear the queue, show execution history, and interrupt a running generation.

Model commands list checkpoints, LoRAs, VAEs and ControlNets, and show details for node types. Image commands list generated images, download one image, or download all images from a prompt. System commands show server statistics (GPU, CPU, memory) and server info including extensions.

Workflow: list, load, validate
Queue: prompt, status, clear, history, interrupt
Models: checkpoints, loras, vaes, controlnets, node-info, list-nodes
Images: list, download, download-all
System: stats, info

Run the skill with natural requests

You do not need to type commands yourself. Describe the result you want and name the ComfyUI setup you have. The agent maps the request to the CLI commands above. Good requests name the workflow file, the output location and whether you want JSON-based results.

Ask for a check first if you are unsure the server is up. The source tells agents to verify ComfyUI with system stats before other commands.

"Check that my local ComfyUI server is running and tell me the GPU memory usage."
"List the checkpoint models available in ComfyUI."
"Queue my_workflow.json and report when it finishes."
"Download all images from prompt ID <id> into ./outputs."

Examples of the commands behind a request

Discover models, queue a workflow, and download results. Use the --json flag when a script or agent needs structured output. Use absolute paths for file operations, as the source recommends.

cli-anything-comfyui models checkpoints
cli-anything-comfyui queue prompt --workflow my_workflow.json
cli-anything-comfyui --json queue history
cli-anything-comfyui images list
cli-anything-comfyui images download-all --prompt-id <id> --output-dir ./outputs

Safety notes: scripts and network access

This skill runs scripts and uses the network. It sends requests to the ComfyUI REST API at localhost:8188. Review the package before installing, and only point it at a ComfyUI server you trust. The skill will queue generation jobs and can interrupt or clear the queue, so ask for a plan before those actions on a shared server.

For agent use, the source asks you to check return codes (0 means success) and parse stderr when a command fails.

Runs scripts: yes. Uses network: yes.
Queue clear and interrupt change the state of a running server.
Output paths are written to the locations you give; use absolute paths.

Tips for reliable results

Keep ComfyUI running before asking for work, and use the --json flag for anything an agent will parse. Use the validate command on a workflow file before queueing it, so errors surface early. Review the history output to find a prompt ID for downloading its images.

Always add --json for parseable output.
Run system stats first to confirm the server is up.
Use history to get the prompt ID for download-all.

What people use it for

Check GPU and memory use on a local ComfyUI server before a large batch.
List checkpoints and LoRAs to see which models are available for a project.
Queue a saved workflow JSON file and monitor it until it completes.
Download every image from a finished prompt into a project output folder.

Frequently asked questions

How do I install the cli-anything-comfyui skill?

Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it in the Claude apps under Settings → Capabilities → Skills. The command-line tool itself installs with pip install cli-anything-comfyui.

What do I need to run cli-anything-comfyui?

You need Python 3.10 or newer, ComfyUI installed and running at http://localhost:8188, and the pip package cli-anything-comfyui. The skill cannot work without a running ComfyUI server.

Is cli-anything-comfyui free?

The source file lists the skill under the Apache-2.0 licence. Check the licence text in the repository for the exact terms before reusing or redistributing it.

Why does the skill say it cannot work without ComfyUI?

The skill talks to ComfyUI's REST API at localhost:8188. If ComfyUI is not installed or not running, every command will fail, so start ComfyUI and check it with system stats first.

Can I use it without a command line or agent?

The skill is entirely command-line. You can run the commands yourself in a terminal, or use the REPL mode, but it is not a visual interface. If you want clicks and a graphical window, use ComfyUI's own interface.

How does it compare to using ComfyUI's graphical interface?

The graphical interface suits building and tweaking workflows by hand. This skill suits automation: queuing workflows from scripts or an agent, reading JSON output, and downloading batches of images.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits creators who already run ComfyUI locally and want to batch-generate or script image jobs, and developers who need a scriptable interface with JSON output. If you do not have ComfyUI running at localhost:8188 or prefer a visual, click-based tool, pick the ComfyUI interface itself instead.

Download the skill