Anthropic has added live iOS app testing to Claude Code Desktop, allowing developers to build, launch, and interact with iOS apps directly within an embedded simulator pane. The feature is available now for Claude users on Pro, Max, and Team plans, and requires macOS with Xcode and the iOS platform installed.
The simulator pane opens alongside the conversation window, letting users watch Claude run an app, interact with its interface, and iterate on changes in real time. Developers can also tap and swipe through the app manually, using the same gestures available in Apple's native Simulator app. The integration means a developer can ask Claude to fix a layout bug, see the result compile and launch in the simulator, and provide feedback on whether the fix worked, all without leaving the chat interface.
The feature is Mac-only for a straightforward reason: Apple's iOS Simulator runs only on macOS. From the desktop app, the simulator pane drives the simulator directly, so it does not need Claude's "computer use" capability and never takes over the user's screen or hides other windows. From the command-line interface, however, Claude reaches the iOS Simulator through computer use instead, controlling it on screen the same way a human would with a mouse.
To use the feature, a developer opens the appropriate tab in Claude Code Desktop, starts a session with their app's project as the working directory, and phrases the task around running or verifying the app. When the app launches, the simulator pane appears next to the conversation. The first time Claude uses a device, the app asks for permission. Claude then installs the app, taps through it, and reads the screen to verify its own changes while the developer watches.
Importantly, the feature is restricted to local sessions. Cloud and SSH sessions cannot access the simulators on a user's Mac, which is a sensible security boundary. The feature requires Claude Desktop version 1.24012.0 or later.
For iOS developers, the addition removes a common friction point in AI-assisted coding. Previously, getting feedback on UI changes meant switching between the IDE, the simulator, and the AI chat window. Now the loop is tighter: write a prompt, watch the code change, see it run, and iterate. It is a small workflow improvement that compounds quickly over a day of development.
The move also reflects a broader arms race among AI coding assistants. GitHub Copilot, Cursor, and Windsurf have all been racing to add deeper IDE and runtime integrations. Anthropic's bet is that controlling the actual execution environment, not just the code editor, is where the real productivity gains live. For mobile developers specifically, being able to see an app render and behave on a simulated device is the difference between theoretical code and working software.