Expo Simulators: Cloud iOS and Android for AI Agents Expo launched cloud iOS Simulators and Android Emulators on November 12, available on all paid plans, giving AI coding agents running on Linux a device runtime to install a build, drive an app, and return a screen recording for verification. Usage costs $0.10 per simulator minute ($6 per hour), and Starter, Production and Enterprise plans include 60 simulator minutes per month until 2027; sessions cap at two hours and the CLI commands are marked experimental at launch. Expo said Agent Sessions, expected before the end of November, will add a secure enterprise layer for triggering agents from external sources with built-in prompt injection protection. Expo launched cloud iOS Simulators and Android Emulators on November 12, now available on all paid plans. The core problem they solve is structural: AI coding agents write mobile code on Linux, but iOS simulators run only on Apple hardware. Agents cannot verify their own work. Expo Simulators https://expo.dev/blog/introducing-expo-simulators-cloud-simulators-built-for-agents close that gap — giving agents a cloud device runtime to install a build, drive the app, and return a screen recording before a human ever has to look at it. The iOS Hardware Lock Was a Real Workflow Problem If you use Claude Code, Codex, or Cursor to write mobile code, you already know the friction: the agent edits your files, but you are the one who has to pull up Xcode, build, and check whether anything broke. That is not an AI workflow — that is just AI-assisted typing with a human verification tax. The fundamental issue is that iOS simulators are locked to macOS. Cloud agent sandboxes run on Linux. Those two facts, in combination, meant your agent was always flying blind. Expo Simulators are built to remove that tax. How It Works The workflow has four stages. Start a session with eas simulator:start via CLI, the REST API, or an authenticated session link. Pass a build — an Expo Build, build fingerprint, application archive URL, or Expo Go. Drive the app using one of three tools: agent-device from Callstack, a CLI with optional MCP integration , Argent from Software Mansion, device control with native MCP , or Appium for teams already on that stack. The session returns a full screen recording, screenshots, and device logs — all readable by the agent and attachable to the pull request. Any agent that can run shell commands can use this. Multiple simulators run in parallel, so an agent testing several screen flows does not have to wait. eas simulator:start --platform ios --build-id