ComfyUI is powerful for image generation but its node-graph API is notoriously complex for AI agents to control directly. ComfyUI Skills for OpenClaw addresses this by wrapping ComfyUI workflows into clean, callable skills that AI agents can invoke with schema-mapped parameters. This turns any ComfyUI workflow into a typed function call that agents like OpenClaw, Claude Code, Codex, and Hermes Agent can use without wrestling with raw node graphs.
How ComfyUI Skills for OpenClaw bridges AI agents and ComfyUI workflows
At its core, this project acts as a bridge layer between ComfyUI workflows and AI agents. It takes workflows exported from ComfyUI in API-format JSON, pairs them with a companion schema.json that defines a clean parameter interface, and exposes these as callable skills.
The main goal is to hide the complexity of ComfyUI’s node graph API from AI agents. Instead of agents manipulating raw node inputs and connections, they control only typed parameters mapped by the schema. This reduces error-prone direct graph manipulation and improves the developer and agent experience.
The repo is implemented in Python 3.10+ and includes a CLI tool named comfyui-skill that supports importing workflows, checking dependencies (like missing nodes or models), running workflows with JSON arguments, and managing multi-server routing to local or remote ComfyUI instances. It also optionally provides a Web UI for visual configuration and testing, which can be helpful for manual debugging and workflow tuning.
While the project supports multiple AI agents — OpenClaw, Claude Code, Codex, Hermes Agent — the core is the abstraction layer and schema-based mapping rather than agent-specific logic. This makes it a practical utility for integrating ComfyUI with agentic workflows.
Technical strengths and tradeoffs: schema mapping and multi-server routing
The standout technical feature is the schema-based parameter mapping that converts ComfyUI’s complex node graph inputs into a clean, typed CLI interface. This is a clear improvement over forcing agents to manipulate node graphs directly, which is error-prone and brittle.
The CLI’s design supports importing workflows along with their schemas and running them with structured JSON arguments. This approach improves automation and integration, especially in multi-agent setups.
Another strength is the multi-server routing capability. This allows the CLI to target either local or remote ComfyUI servers seamlessly, which is practical in distributed or cloud workflows.
The codebase is Pythonic and leverages modern Python features. The CLI tooling is fairly minimal but covers essential commands for workflow management and execution.
Tradeoffs include the fact that the abstraction only goes so far: agents still need to understand the parameter schema to use the skills effectively. The underlying complexity of ComfyUI workflows and node graphs remains hidden but not eliminated.
Also, the project assumes users have a running ComfyUI server and exported workflows in API JSON format, so it doesn’t handle workflow creation or editing itself.
Overall, the code quality appears solid with clear separation of concerns between workflow import, dependency checking, and execution.
Quick start with comfyui-skill CLI
Get ComfyUI Skills running in a few minutes.
Before you start, make sure you have:
- Python 3.10+
- A running ComfyUI server
- An exported workflow in ComfyUI API format
1. Clone the project
Choose the directory that matches your agent environment.
For OpenClaw
cd ~/.openclaw/workspace/skills
git clone https://github.com/HuangYuChuh/ComfyUI_Skills_OpenClaw.git comfyui-skill-openclaw
cd comfyui-skill-openclaw
For Claude Code
cd ~/.claude/skills
git clone https://github.com/HuangYuChuh/ComfyUI_Skills_OpenClaw.git comfyui-skill
cd comfyui-skill
For Codex
cd ~/.codex/skills
git clone https://github.com/HuangYuChuh/ComfyUI_Skills_OpenClaw.git comfyui-skill
cd comfyui-skill
For Hermes Agent
cd ~/.hermes/skills/creative
git clone https://github.com/HuangYuChuh/ComfyUI_Skills_OpenClaw.git comfyui-skill-openclaw
cd comfyui-skill-openclaw
Or install via Hermes CLI (once the PR is merged):
hermes skills install comfyui-skill-openclaw
2. Create your local config
cp config.example.json config.json
3. Install the CLI
pipx install comfyui-skill-cli
Or:
pip install comfyui-skill-cli
If you already have the CLI installed, upgrade it with:
# If you installed it with pipx
pipx upgrade comfyui-skill-cli
# If you installed it with pip
python3 -m pip install -U comfyui-skill-cli
4. Verify the setup
comfyui-skill server status
comfyui-skill list
5. Import and run your first workflow
comfyui-skill workflow import /absolute/path/to/my-workflow.json
comfyui-skill deps check local/my-workflow
comfyui-skill run local/my-workflow --args '{"prompt": "a white cat"}'
This workflow import expects an absolute path to avoid ambiguity. From there, you can check dependencies and run the workflow with JSON arguments specifying parameters exposed by the schema.
Verdict
ComfyUI Skills for OpenClaw is a practical tool for anyone integrating ComfyUI image generation workflows into AI agents or automation pipelines. Its schema-based abstraction provides a cleaner and safer way to invoke complex workflows without dealing directly with node graph internals.
However, it’s not a full workflow editor or a replacement for understanding ComfyUI’s graph logic. Agents and developers still need to understand and manage parameter schemas, and the underlying ComfyUI server and workflows must be prepared separately.
If you’re building AI agents that need to control image generation workflows programmatically, this repo offers a solid foundation for wrapping ComfyUI workflows into callable skills. It’s especially relevant if you use OpenClaw, Claude Code, Codex, or Hermes Agent environments.
The multi-server routing and dependency checks add useful production touches, although expect to invest some time in understanding your schemas and workflow exports before full automation is smooth.
Overall, the project addresses a real pain point in agentic control of ComfyUI and does so with a clean, Pythonic CLI interface and reasonable tooling.
→ GitHub Repo: HuangYuChuh/ComfyUI_Skills_OpenClaw ⭐ 293 · Python