Visual output surface for agents: render and preview images and videos from text, image, or embedding inputs. Starts with a ComfyUI backend but is not locked to it — pluggable generation backends behind one agent-first CLI.