What is comfy-agent-tools?

comfy-agent-tools is a Python project that gives AI agents a set of skills and command line tools for generating images, videos, and music locally on your machine, using a library called comfy-diffusion as its core engine. Instead of requiring a server with a web interface, it calls generation pipelines directly from the command line, keeping things lightweight. The typical workflow starts with installing agent skills, which are small instruction sets that teach your AI coding assistant how to set up, update, and run the tools automatically. Once installed, you can ask your agent something like set up comfy-agent-tools and generate an image, and it handles the rest: downloading missing model files, initializing configuration, and running the generation command. The individual tools, such as comfy-imagegen, comfy-videogen, comfy-musicgen, and others, map to specific creative capabilities: still images including editing and upscaling, video clips, and music files in WAV format. Model files for local generation live in a directory you configure, and are downloaded on demand only when actually needed, not all at once upfront. Some capabilities use remote APIs with provider credentials instead of downloading large local weights. Profiles let you map each capability to a specific model configuration, and you can extend or override the built-in defaults with your own fine tuned checkpoints. The project is written in Python and uses uv as its package and tool installer. It supports both locally run AI models and remote API based generation for select capabilities. The full README is longer than what was shown.