Claude Code Now Runs Your iOS App in the Simulator While You Watch

Claude Code Desktop can now build, run, and test an iOS app in Apple's Simulator inside a pane next to the conversation, closing a loop that used to force developers back and forth between Xcode and the CLI. Kevin reacts to a July 21, 2026 post from the @ClaudeDevs account announcing the feature, which shipped in public beta. The screenshot he reacts to shows a split view: a Claude Code session on the left and a live iOS simulator on the right running a plant-care app called Soyary. His main point is that the whole build-run-test cycle is now agentic instead of manual.

What the announcement actually says

The frames show the @ClaudeDevs post front and center: "Claude Code on desktop now works with the iOS simulator. Build and run your iOS app, and the simulator opens in a panel right next to your conversation. Available today in public beta." A reply visible below adds that Claude "can see the app as it runs, interact with it, and iterate until finished," and that you can "tap through the simulator yourself at any time." The attached screenshot pairs a Claude Code conversation with a booted iPhone running the Soyary app, cards for Monstera, Pothos, and Fiddle leaf visible on screen, and a "Claude is using this device" badge above the phone.

This matches Anthropic's own documentation. The iOS Simulator pane is in public beta in Claude Code Desktop on macOS, available on Pro, Max, and Team plans. When Claude builds, installs, launches, or checks an app in a simulator, the pane opens automatically and streams the device screen live. Up to four simulator panes can run per session.

Why the old workflow was painful

Kevin frames the value against how iOS development with AI worked before. As he puts it in the transcript, "even six months ago, you had to literally go back and forth between Xcode and your CLI. And sometimes if you plug in your CLI into Xcode, it really didn't work that well." The pattern was: operate in Claude Code, switch to the CLI, check how it worked, simulate it yourself, test things, then go back. Every verification step was a manual context switch.

He also notes that running Claude Code against an iOS app in the Xcode simulator was already possible before this, but only through "a few GitHub repos or a few different open source tools." Open-source MCP servers like XcodeBuildMCP had been filling that gap, giving agents the ability to build, boot a simulator, install a binary, tap through the UI, and read logs. The TikTok's point is that this capability is now built into Claude Code Desktop directly rather than stitched together from third-party tools.

What "agentic" means here in practice

The distinction Kevin lands on is that "now this entire process is agentic." Per Anthropic's docs, the pane drives the simulator directly, so it does not rely on computer use and never takes over your screen or hides other windows. You phrase a task around running or verifying the app, for example "build the app and run it in the simulator to check the onboarding flow," and Claude installs the app, taps through it, and reads the screen to verify its own changes while you watch. You and Claude drive the same device, so you can tap through a screen yourself and then ask Claude about it. The "Claude is using this device" badge visible in the frames signals when Claude is the one controlling the phone.

The requirements and the catch

The feature is macOS only, because Apple's iOS Simulator runs only on macOS. It needs Claude Desktop v1.24012.0 or later and Xcode with the iOS platform installed to provide the simulator devices. The pane works in local sessions only, not cloud or SSH sessions, since those run on a machine that cannot reach the simulators on your Mac. Anthropic also flags a privacy note: Claude's screenshots of the device are sent to Anthropic under normal conversation retention, so you should not sign in to real accounts on a device Claude uses.

Where Kevin lands

His closing reaction is telling for the target audience. He says the feature "makes me just want to build some apps just for personal use, like not even publish them on the app store, just have them loaded on my phone." That is the practical unlock: lowering the friction of iOS development enough that shipping a personal app for your own phone becomes a casual project rather than a multi-tool ordeal. The video also references Google Gemini building simple Android apps from a single prompt in its description, positioning both as evidence that app building has gotten dramatically easier.

Key Takeaways

Claude Code Desktop can build, run, and test an iOS app in a live simulator pane next to the conversation, in public beta on macOS.
The pane drives the simulator directly without computer use, so it never takes over your screen.
You and Claude share the same device: Claude taps through and verifies its own changes while you can tap through yourself.
Requirements are macOS, Claude Desktop v1.24012.0 or later, and Xcode with the iOS platform installed. Local sessions only.
Before this feature, running agents against the iOS simulator required third-party open-source tools like XcodeBuildMCP.
Do not sign in to real accounts on a simulator device Claude uses, since screenshots are sent to Anthropic.