This project provides an agent-friendly interface for ComfyUI workflows, enabling AI agents like OpenClaw, Hermes, Codex, and Claude Code to execute them via a Command Line Interface (CLI). It transforms complex ComfyUI graphs into callable skills with schema-based parameter mapping, offering a controlled and reusable way to integrate generative AI workflows into agent tasks. An optional Web UI aids in configuration and testing.

How It Works

The core mechanism involves wrapping ComfyUI workflows with a schema-driven layer. This layer exposes only essential parameters, defined with clear types and descriptions, allowing agents to interact predictably via shell commands. The project automatically imports ComfyUI workflow JSON files, generates these schema mappings, and provides a CLI for workflow execution, dependency checking, and multi-server management. This approach abstracts away the intricacies of ComfyUI's graph structure, presenting a clean, agent-consumable API.

Quick Start & Requirements

• Primary Install: pipx install comfyui-skill-cli or pip install comfyui-skill-cli. Clone the repository into the appropriate agent workspace directory.
• Prerequisites: Python 3.10+, a running ComfyUI server instance, and exported ComfyUI workflows in API format.
• Setup: Configure config.json with ComfyUI server details. Import workflows using comfyui-skill workflow import /path/to/workflow.json.
• Links: Demo Video, Docs, Quick Start, CLI, Web UI.

Highlighted Details

• Agent-Friendly CLI: Designed for programmatic interaction, offering cleaner interfaces than direct ComfyUI graph manipulation.
• Schema-Based Parameter Mapping: Exposes configurable parameters with aliases, types, and descriptions for controlled agent input.
• Multi-Server Routing: Manages and routes jobs across multiple local or remote ComfyUI servers under a single namespace.
• Dependency Management: CLI can check for and assist in installing missing ComfyUI nodes and models required by workflows.

Maintenance & Community

The project has seen recent updates, including a migration to a CLI-first architecture (v0.4.0) and added support for API keys (v0.3.1) and dependency management (v0.3.0). Contribution guidelines are available in CONTRIBUTING.md. Links are provided to the frontend repository, Hermes Agent, and the agentskills.io standard. Specific community channels (Discord/Slack) or maintainer details are not explicitly listed.

Licensing & Compatibility

The project's license is not explicitly stated in the README, which may pose a barrier for commercial adoption or integration into closed-source systems. It is designed for compatibility with AI agents capable of executing shell commands and adheres to the agentskills.io open standard.

Limitations & Caveats

Workflows must be exported in ComfyUI API format and include an output node (e.g., Save Image). A schema.json file is required for parameter mapping. The target ComfyUI server must have all necessary custom nodes and models installed; the CLI can assist in checking these. Invalid payloads or incorrect schema mappings can lead to HTTP 400 errors. The lack of a specified license requires clarification for adoption decisions.