Anthropic shipped iOS Simulator support in Claude Code Desktop on July 21 as a public beta — and it works differently than you’d expect. The simulator doesn’t open in a separate window you alt-tab to. It opens as a pane directly beside your Claude conversation, streams live, and lets Claude build, launch, test, and iterate on your app without touching your desktop or demanding Accessibility permissions. For iOS developers already using Claude Code, this closes the one gap that forced constant tool-switching.
The Pane, Not the Window
The iOS Simulator pane appears inside Claude Code Desktop’s Code tab, adjacent to your conversation. When Claude builds, installs, launches, or checks your app in a simulator, the pane opens automatically and streams the device screen live. You don’t configure it — it just appears when the build succeeds.
You can watch Claude run and test the app, or tap through the app yourself while Claude keeps working. Hardware shortcuts work as expected: Cmd+Shift+H for Home, Cmd+L to lock, Cmd+Up and Down for volume. Each session gets its own simulator, and you can have up to four panes open simultaneously — useful if you need to compare behavior on an iPhone SE and an iPhone 16 Pro Max in the same session. To target a specific device, name it in your prompt: “run it on the iPhone SE simulator.”
Why This Is Not Computer Use
This distinction matters more than it sounds. Claude’s existing computer use feature controls your screen through macOS Accessibility and Screen Recording permissions — it literally takes over your cursor and watches pixels. That’s why many developers never enabled it: granting full screen access to an AI agent is a non-trivial security concession, and having the agent take over your desktop mid-session is disruptive.
The iOS Simulator pane takes a different approach. It communicates directly with Apple’s Simulator APIs rather than controlling screen pixels. No Accessibility permission. No Screen Recording. No screen takeover. Your other windows stay exactly where they are. The agent works in its own lane.
This is the architectural move that makes the feature practical. A tool that works without demanding sensitive system permissions is a tool developers will actually use.
Setup: What You Actually Need
The requirements are reasonable but specific:
- macOS only. Apple’s iOS Simulator is macOS-exclusive. There is no Windows or Linux path here.
- Xcode with the iOS platform installed. Standard for any iOS developer, but worth confirming it’s not just the Xcode Command Line Tools.
- Claude Code Desktop v1.24012.0 or later. Update if you haven’t recently — check the official docs for the latest minimum version.
- Local session. Cloud and SSH sessions run on a remote machine that can’t reach the simulators on your Mac. This feature requires local.
- Pro, Max, or Team plan. Enterprise is excluded — more on that below.
To start, open the Code tab in Claude Code Desktop, set your app’s project folder (a .xcodeproj or .xcworkspace directory), and phrase your task around building or running the app: “Build the app and run it in the simulator to check the onboarding flow.” The pane opens when the build succeeds.
The Enterprise Gap
Enterprise plan users don’t get this feature. That’s the most notable limitation in an otherwise clean release. Teams on Anthropic’s Enterprise plan — the one most likely to have iOS development workflows worth automating — are sitting this beta out. Anthropic hasn’t explained the exclusion, though it likely involves the privacy model: simulator screenshots are sent to Anthropic under your conversation retention settings. Enterprise customers often carry stricter data handling requirements, and a beta that transmits screen content may not meet those standards yet.
For individual developers and small studios on Pro or Max, none of this applies. For a team lead evaluating whether to roll this out across an iOS shop, the Enterprise exclusion is the first question to answer.
What This Changes (and What It Doesn’t)
Before this, Claude Code could write Swift and run builds, but it couldn’t see the result. Developers had to check the Simulator manually, describe what was wrong in a follow-up prompt, and loop back. Every UI bug required a human relay. The simulator pane removes that relay: Claude writes code, builds, launches, sees the running app, spots a broken layout or crash, and fixes it — without a prompt from you.
That said, this isn’t fully autonomous. You stay in the loop. You can tap and swipe in the pane while Claude works, redirect the agent mid-session, or take over entirely. The hybrid model is intentional — Anthropic isn’t shipping a “build my app without me” feature, it’s shipping a tool that makes the AI-assisted development loop tighter and faster without removing the developer from it.
One practical note on context costs: each simulator screenshot goes into Claude’s context window. Long sessions or sessions with frequent visual checks will consume more tokens. Keep your prompts focused if you’re watching costs.
The iOS Simulator pane is in public beta, which means Anthropic can change behavior, adjust plan access, or pull it without notice. Early reactions from the community are positive, but treat it as a preview and plan accordingly. If you’re building iOS features with Claude Code today, it’s worth enabling — with the caveat that the beta label is real. For more on the broader agentic iOS development picture, our Xcode 27 agentic coding guide covers how MCP and on-device AI fit into the same workflow.
