The simulator joins the conversation

Building an iOS app usually means moving between the code, the build output and a separate simulator view. Anthropic’s release draws those jobs together: the simulator opens beside the Claude Code conversation when we build and run the app.

Anthropic says Claude can see and interact with the running app while we remain free to tap through it ourselves. That creates a shared test loop, although the announcement does not establish faster work, greater reliability or a production-ready result.

The practical change is continuity. We can ask for an edit, build the app and inspect the result without losing sight of the conversation that produced it. Claude can observe the same running interface and respond to what appears there.

That does not remove judgement from the process. A visible app state can expose an obvious layout or interaction problem, but it cannot prove that the code is correct, complete or ready to ship.

Anthropic says the capability is available today in public beta. That wording matters: it describes an early public release, not general availability or established production performance.

Trying it requires Claude Code on desktop, a Mac and Xcode. The two announcement posts make no claim about Windows, Linux, physical devices, pricing or support beyond the iOS simulator.

The announcement goes beyond placing a simulator panel beside the conversation. Anthropic says Claude can observe the app as it runs and interact with its interface.

We can also tap through the simulator at any time. The source therefore describes shared access to the running app, not an autonomous hand-off of development or verification.

Claude can see the app as it runs, interact with it, and iterate until finished. Tap through the simulator yourself at any time.

ClaudeDevs on X

The approved evidence consists of two official ClaudeDevs posts and one announcement image. Together they establish the intended workflow, public-beta status and stated platform requirements.

They do not tell us whether the loop is faster, more reliable or more accurate in practice. They also do not establish test coverage, security, pricing, physical-device support, deployment capability or App Store readiness.

The loop starts with a change inside the current project. Claude Code can work on that change, build and run the app, then use the simulator displayed beside the conversation to observe what happened.

From there, Claude may interact and iterate, while we can step in at any point. The useful distinction is between assisting with a visible test loop and proving the finished app: the announcement supports the former, not the latter.

Anthropic’s wording keeps direct human access explicit: we can tap through the simulator at any time. That makes the feature useful as a collaborative inspection surface rather than a claim of unattended app delivery.

The setup still depends on macOS and Xcode, and the result still needs our scrutiny. A successful simulator run can show what happened in that session; it cannot settle every device, reliability or release question.

The meaningful change is shared visibility: Claude can work with the running app beside the conversation, while we remain responsible for deciding whether the result holds up.

Altior analysis of Anthropic’s announcement

Anthropic has connected Claude Code’s conversation to a live iOS simulator on macOS. That could make the path from request to visible result feel more continuous because the code, running interface and discussion occupy the same workspace.

The honest conclusion is narrower than a promise of autonomous app creation. This public beta gives us a new way to observe and steer iteration; whether it improves speed, quality or reliability still needs to be demonstrated in use.