Computer use traces show up in the Gemini desktop app

Google has begun testing computer use in the Gemini desktop app, a mode that would let the assistant control other applications and open files in folders the user selects. Traces of the feature turned up in the app's settings, and Testing Catalog reports that this suggests testing with a limited group of trusted testers.

At a glance

• The feature is expected to surface as a "Task" option in the prompt-bar selector, sitting next to the existing controls for image and video generation rather than as a toggle in the navigation bar.
• While Gemini works, a mini-view of the active screen shows what the agent is doing, and a separate control keeps the machine from going to sleep while a task is still running.
• An optional setting, described by Testing Catalog as Advanced Backup, copies the selected folders to Google Drive before Gemini starts modifying files, a recovery path for users without Git or another version-control system.

Computer use reads as Google's next major frontier for Gemini, and it is the capability that most clearly separates a chat assistant from an agent with hands on the file system. The presence of an automatic backup switch is itself a signal: it implies the agent can make changes a user would want undone. Value will likely hinge less on what the agent can do than on permissions, recovery and oversight.

Gemini would control other applications and open files in user-selected folders

The functionality would let Gemini operate other applications on the machine and open files in folders the user has designated, covering the kind of desktop work that Claude and ChatGPT already handle. The Gemini desktop app has been under development for some time, and computer use is the capability that had not previously surfaced in it.

Evidence of the feature comes from the app's settings, where the relevant traces appeared. According to Testing Catalog, that placement suggests Google has begun testing with a limited group of trusted testers rather than exposing the mode to the general desktop audience.

The "Task" entry in the prompt-bar selector replaces a navigation-bar toggle

Computer use is expected to ship alongside a slightly updated Gemini Spark interface. Rather than being switched on through a toggle in the navigation bar, it is expected to appear as a "Task" entry in the prompt-bar selector, the same control that already holds image and video generation.

While the agent is driving the machine, users will be able to follow its actions through a mini-view of the active screen. The app is also expected to carry practical controls around long-running jobs, including one that stops the device from going to sleep while a task is in progress.

Advanced Backup copies selected folders to Google Drive before Gemini edits them

An optional safety feature automatically backs up selected folders to Google Drive before Gemini begins a task. Testing Catalog describes it as an Advanced Backup option: Gemini stores copies of the files on Drive first, and only then modifies the originals in the selected folders.

For users who do not keep their work in Git or another version-control system, the copy on Drive becomes the rollback path after an agent-driven edit goes wrong. Because the backup is optional, it has to be enabled before a run for that copy to exist when Gemini starts changing files on the machine.

What happens with hundred-gigabyte folders

Google has not announced a release date for computer use in the Gemini desktop app, and the settings traces do not indicate how wide the trusted-tester group is. The backup mechanism also leaves a practical question open: a folder holding hundreds of gigabytes would take a long time to upload to Google Drive and would consume substantial storage before the task could start.