The project page describes three complementary components, including VideoCUA, UI-Vision, and GroundCUA. That structure suggests a platform that combines demonstrations, annotation, and grounding data to support robust agent learning. It is especially valuable for researchers and builders who need diverse, labeled interaction traces.
CUA-Suite stands out because it focuses on real human demonstrations across a wide range of applications. That makes it a strong foundation for computer-use agent research, benchmarking, and data-driven training.
