Claude Code Can Now See Your Screen, Click Your Buttons, and Test Its Own Work
Anthropic shipped 'computer use' inside Claude Code, letting the AI open apps, interact with UIs, and verify its output — all from the command line. It's the most aggressive agentic capability any lab has put in developers' hands, and it's already hitting scaling walls.
Anthropic released computer use inside Claude Code on Monday, giving the AI the ability to open applications, click through graphical interfaces, and test the software it writes — all without leaving the CLI. The feature, available as a research preview for Pro and Max plan subscribers, represents the most ambitious agentic deployment any major lab has shipped to a general developer audience. As @claudeai announced, Claude can now "open your apps, click through your UI, and test what it built, right from the CLI."
The capability closes a loop that has frustrated AI-assisted development for years. Previously, coding agents could write code and run it, but verifying that a UI actually rendered correctly, that a button was clickable, or that a workflow completed as designed required a human in the loop. Computer use eliminates that bottleneck. As @shota7180 documented, the feature enables a full pipeline: code creation, compilation, app launch, UI manipulation, bug detection, fix, and verification — all orchestrated by the model.
Get our free daily newsletter
Get this article free — plus the lead story every day — delivered to your inbox.
Want every article and the full archive? Upgrade anytime.
No spam. Unsubscribe anytime.