Claude Code on desktop now works with the iOS simulator

I'll pull the official docs page you linked so the paragraphs stick to real product details and avoid invented numbers or claims.Claude Code Desktop on macOS now includes an iOS Simulator pane that opens when Claude builds, runs, installs, launches, or checks an iOS app. The feature is in public beta on Pro, Max, and Team plans and is not available on Enterprise. It requires Claude Desktop v1.24012.0 or later, a Mac, and Xcode with the iOS platform installed so Apple simulator devices are available. Coverage on Hacker News is still thin (1 point, 0 comments) at the linked thread, with the primary source being Anthropic’s own desktop docs for the pane.

Technically, the pane sits beside the conversation and streams the simulated iPhone or iPad screen live. Claude Code Desktop drives the simulator directly, so it does not depend on general computer-use screen control and does not take over or hide other windows. From the CLI, the same class of work still goes through computer use and operates the Simulator app on-screen the way a mouse would. The pane is interactive: taps, swipes, Home, lock, volume, rotate, screenshots (Cmd+S), and recordings (Cmd+R) work from the pane, with stream tunables for frame rate, resolution, and H.264 versus JPEG encoding. Each device is owned by the session that launched it, so parallel sessions do not share a device; a session can open up to four simulator panes. Claude-booted simulators shut down when you quit the app, archive the session, or ten minutes after you detach; devices you boot yourself are never shut down automatically. First-time Claude control of a device needs a one-time consent that covers control and screenshots; those screenshots go to Anthropic under normal conversation retention. Build and open-URL actions still follow the session permission mode because they run local build scripts or can move data off-device.

For engineers building iOS apps with an agent in the loop, the useful shift is closed-loop UI verification without leaving the desktop session. You can ask Claude to build and run an onboarding or signup flow, watch install and launch in the pane, then either let Claude tap and read the screen or navigate yourself and ask it to inspect the current state. The “Claude is using this device” badge makes shared control explicit so human taps do not corrupt Claude’s verification. Naming a device (for example iPhone SE) targets that simulator; Claude can also install onto a device you already have booted in Apple’s Simulator app. Because the pane only works in local sessions, cloud and SSH sessions cannot reach the simulators on your Mac—remote agent runs still cannot close that loop on your machine’s simulator stack.

In product terms this sits between generic screen-control agents and IDE-adjacent agent tooling. Dedicated simulator streaming avoids Accessibility and Screen Recording permissions that full computer use needs, while still tying into Xcode’s device list and xcodebuild. That is a narrower, safer surface for mobile UI work than grabbing the whole desktop. Competitors in coding agents often rely on screenshots, CLI build logs, or general computer use for simulators; a first-class pane with per-session device isolation and plan-gated availability is a desktop-specific differentiation for Claude Code rather than a CLI-only trick. Physical devices remain out of scope: Claude cannot control a real iPhone or iPad; on-device checks still mean running from Xcode yourself and describing or attaching a screenshot.

Practical next steps if you ship iOS: confirm Desktop is at least v1.24012.0, install an iOS simulator runtime if the pane reports none found (Xcode settings or `xcodebuild -downloadPlatform iOS`), and phrase tasks around run/test so the pane auto-opens. Use the Views menu Attach simulator path when you want the pane without waiting for a launch. Do not sign into real accounts on a device Claude can screenshot. Watch whether Enterprise gets the feature later, whether org managed settings (`disableMobileSimulatorTools` or VM-isolated sessions) block it in your shop, and whether physical-device or remote-session support ever lands; until then, treat this as a local macOS beta loop for simulator-only UI validation inside Claude Code Desktop.