Claude Code adds computer use in research preview for Pro and Max

• Anthropic's launch post puts computer use into Claude Code, so the CLI can now "open your apps" and "click through your UI" while testing what it built.
• In Anthropic's own workflow detail, the new loop can happen in one prompt: write code, compile it, launch the app, click through it, then "find the bug, fix it, and verify the fix."
• Availability is limited for now: Anthropic's launch post says the feature is in research preview on Pro and Max plans, and a supporting report from Wes Roth says the rollout is currently macOS-only.
• Early community posts such as TestingCatalog's demo frame this less as a new model and more as a workflow change: Claude Code can now operate the local app it is building against from the terminal.

Anthropic added its Computer Use capability directly to Claude Code, extending the CLI from code generation into UI execution. The company says Claude can "open your apps, click through your UI, and test what it built" from the terminal, and the attached demo launch demo shows a prompt triggering an app launch and an automated button click.

The bigger change is that Anthropic is now describing a closed-loop GUI debugging workflow, not just command execution. In Anthropic's workflow detail, Claude can in one pass "write the code, compile it, launch the app, click through it, find the bug, fix it, and verify the fix," which effectively moves basic end-to-end validation into the same prompt that generated the feature.

A supporting demo from TestingCatalog demo restates the practical shift: Claude Code can "operate apps that you are building." For engineers, that means the agent is no longer limited to source files and shell output; it can act against the running local interface it just produced.

Anthropic's plan availability says the feature is live in research preview for Pro and Max plans. That makes this a paid-tier rollout rather than a general Claude Code default, and it signals that Anthropic is still treating the capability as an early access feature rather than a stable production surface.

A supporting post from Wes Roth report adds a platform constraint not stated in the launch tweet: the rollout is "strictly for macOS users" at this stage. If that holds, the preview is narrower than the headline suggests — both by subscription tier and by OS.

What the preview appears to cover is direct desktop control from the terminal. The same report describes Claude as able to "physically move the mouse, click through" local GUIs and verify on-screen actions, which aligns with Anthropic's own description of app launching and UI interaction rather than a browser-only or sandboxed test flow.