comfyui

The ComfyUI skill provides a robust interface for OpenClaw to interact with a self-hosted ComfyUI server running on your local network. By leveraging the power of your own hardware, this skill ensures total privacy for your creative process while providing access to complex, node-based generative AI workflows. It is specifically designed for users who require professional-grade image synthesis, custom SDXL model support, and the flexibility of managing their own image generation pipelines without relying on external cloud-based services.

• Full integration with local ComfyUI servers via WebSocket and HTTP API protocols.

• Support for custom JSON workflow files, allowing for complex node setups, LoRA weight adjustments, and specific sampling configurations.

• Default optimization for SDXL models like Juggernaut XL to ensure high-fidelity output.

• Automated image management capabilities, including saving results to local directories and optional synchronization with document folders.

• API-first architecture, enabling seamless programmatic triggering of image generation directly from the agent terminal.

• Requires a running ComfyUI instance with Dev mode enabled to accept API requests.

• Configure the COMFYUI_SERVER_ADDRESS environment variable to point to your instance (e.g., http://192.168.x.x:8189).

• Use the provided comfy_gen.py script to queue prompts, process custom workflows, and manage image output files.

• If workflow paths contain special characters, ensure they are enclosed in single quotes during command execution.

• Note that LoRA weight modifications should be managed directly within your workflow JSON files as the script defaults to prompt injection and seed randomization.

• Ideal for creative professionals, researchers, and engineers who need reliable, offline, and repeatable image generation pipelines.