Gemini API Computer Use goes generally available for browsers and desktop apps

The Computer Use tool in the Gemini API has moved from the experimental stage to general availability. The change allows agents built on Google’s platform to control browsers and desktop applications, instead of only processing text or calling pre-structured APIs.

The core of Computer Use is the ability to operate an interface like a user. Agents can work with software that does not provide APIs, which accounts for a large share of the systems running in enterprises, thereby extending automation to tasks that previously had to be done manually.

As described, the control is exercised with safeguards. This factor matters because an agent acting directly on the screen can reach data, accounts and actions that cannot be undone.

A computer-control tool reaching general availability indicates that agent development is shifting from conversational assistant to executing actor. Even so, operational safety still depends on how the deploying organisation limits permissions, logs activity and approves critical steps.

Source: ai.google.dev