Claude Code’s Desktop App Adds iOS Simulator Support: Run Your App Beside the Conversation

July 22, 2026Updated: July 25, 2026Articles

On July 21, 2026, Anthropic launched a public beta that brings the iOS Simulator directly into the desktop version of Claude Code.

Claude Code could already write code for an iOS app, but checking the result often meant moving to Xcode or Simulator and inspecting the build yourself. With this update, an iPhone or iPad simulator opens beside the Claude Code conversation. You can watch Claude build and launch the app, then see it interact with the interface.

For iOS developers, the important change is that the workflow no longer has to stop when the AI says it has written the code. The running interface is now much closer to the conversation where the work began.

The iOS app runs right beside the conversation

The new iOS Simulator pane appears in the Code tab of the Claude Code desktop app. Open a project and ask something such as, “Create an onboarding screen, run it in the simulator, and check the result.” Claude can build, install, and launch the app, opening the simulator automatically when it is needed.

The pane is not just a video feed. You can interact with it too.

Click and drag to test taps and swipes
Rotate the device
Switch the simulated device or OS version
Save screenshots and screen recordings
Navigate to a particular screen yourself, then ask Claude to inspect it

You and Claude see the same simulator. That means you can navigate to the screen where a problem occurs and say, “The button is clipped in this state—please fix it,” without first taking a screenshot and attaching it to the conversation.

The real improvement is visual verification, not code generation

In AI-assisted development, successfully generating code is not the same as creating a usable app. A build can pass while text is cut off, buttons are difficult to tap, or navigation behaves differently from the design.

This feature brings that final check into Claude Code. It makes requests like these easier:

Check whether the layout breaks on a smaller iPhone
Walk through the flow from first launch to completed registration
Repeat the same interaction after a fix to confirm the bug is gone
Compare the same screen across several device sizes

In other words, this is not only a way to delegate implementation to Claude. It shortens the loop from build to run, inspect, and revise. That could be particularly helpful for people who are not deeply familiar with code but want to build a small iOS app for themselves without constantly switching between Xcode and the AI conversation.

Requirements to check before you start

It is not available to every user or in every environment. As of July 22, 2026, Anthropic’s documentation listed these requirements:

A Mac
Claude Desktop version 1.24012.0 or later
Xcode, the iOS platform, and at least one simulator device installed
A local Claude Code session
An eligible Pro, Max, or Team plan

Cloud and SSH sessions cannot connect to Simulator on your local Mac. Enterprise plans were also excluded when the beta launched.

If your environment meets the requirements, you do not need to learn a special command. Open an iOS project in Claude Code’s desktop app and clearly state the goal—for example, “Run this in iOS Simulator and check this screen.”

You do not need to enable Computer Use

The simulator pane is separate from Computer Use, where Claude controls the broader Mac interface. It talks directly to Apple’s simulator tooling, so it does not require giving Claude macOS Accessibility or Screen Recording permissions. It also does not hide other windows or take over the whole screen while working.

The project’s build scripts still run on your Mac, however. Opening a URL in the simulator can also send information to an external service. Both operations follow the permission mode selected in Claude Code.

It cannot control a physical iPhone, and the beta has limits

The feature supports only virtual iPhone and iPad devices provided by Xcode. Claude cannot control a physical iPhone connected by USB or Wi-Fi. Cameras, notifications, Bluetooth, real-world network conditions, and device-specific performance still need to be tested on real hardware.

A public beta is mature enough for ordinary users to try, but its interface, eligible plans, and stability may change. Start with a small personal app or test project instead of immediately relying on it for all validation of an important product.

Do not sign a simulator into your real accounts

The first time Claude tries to control a simulator device, it asks for permission for that device. Once approved, Claude can tap, enter text, launch apps, and capture screenshots.

Anthropic’s documentation explains that simulator screenshots captured by Claude are sent to Anthropic and handled under the same retention settings as ordinary conversations. Do not sign a test device into your personal Apple Account or everyday services. Use dummy names, test email addresses, and test data.

Start with one small screen

Rather than asking Claude to build an entire app first, use a single screen whose result is easy to judge. That makes the strengths and limitations of the feature clearer.

Ask for a simple screen containing only a button and input field
Launch it in iOS Simulator and have Claude inspect the layout
Tap it yourself and describe anything that feels wrong
Check the same device again after the revision
Finally, try another device size and look for layout problems

iOS development has not become fully automatic just because Claude Code can access iOS Simulator. The distance between AI-generated code and human visual inspection has, however, become much shorter.

If you already have a Mac and Xcode, try it first on an existing small project with a request such as, “Run this screen in the simulator and inspect it.” That is enough to see whether the update fits your way of working.

References

Features and availability were checked on July 22, 2026. The author had not tested the public beta on physical hardware at the time of writing; the instructions and limitations above are based on Anthropic’s official announcement and documentation.