Computer use
A model driving a screen directly — looking at pixels, moving a cursor, typing — instead of calling an API.
Also written: browser use, GUI agent.
Where no API exists, a model can be given a screenshot and the ability to click and type. It looks at the screen, decides where to click, and does it.
It is slow, expensive in image tokens, and considerably less reliable than a real integration. It is also the only option for the long tail of software that has no API, which is most software.