FAQ
Frequently asked questions about ComfyUI Skills for OpenClaw
These answers are written for users comparing ComfyUI integrations, OpenClaw skills, and agent-friendly workflow tooling.
What is ComfyUI Skills for OpenClaw?
ComfyUI Skills for OpenClaw is an open-source bridge that turns ComfyUI API-format workflows into callable skills for OpenClaw and compatible AI agents.
What problem does this repository solve?
It lets an agent take natural-language requests, map them to workflow parameters, submit jobs to one or more ComfyUI servers, wait for completion, and download the generated images automatically.
Can this project work with multiple ComfyUI servers?
Yes. The repository supports multi-server workflow namespaces, so one OpenClaw setup can route jobs across different ComfyUI servers instead of being locked to a single machine.
What files are required for a workflow to work?
Each workflow needs an API-format workflow.json exported from ComfyUI and a schema.json file that exposes the parameters the agent is allowed to call.
Does the project include a web UI?
Yes. The repository includes a local web dashboard for uploading workflows, configuring servers, and editing schema mappings.
Is this only for OpenClaw?
OpenClaw is the main target, but the architecture is also useful for other skill-capable AI agents that can call structured tools or local skills.
Why does the repository use schema.json instead of exposing the whole graph?
The schema layer gives the agent a smaller and safer contract with clear aliases, types, descriptions, and required flags, which makes workflow calls more reliable than exposing every node field directly.
What is the recommended way to configure the first workflow?
The recommended flow is to use the local web dashboard first, because it lets you upload an API-format workflow, inspect the parsed inputs, and choose which parameters should be exposed before saving.
How do I know a workflow export is valid for this project?
The workflow should be exported from ComfyUI using Save API Format, and it should include a Save Image style output so the client can download generated images after the job completes.
Can I move my workflow mappings to another machine?
Yes. The repository includes export and import flows for transferring server definitions and workflow mappings, which is useful when moving a skill setup to another machine or workspace.