Open-source iPhone Use lets Codex control a real iPhone

Former Doubao desktop product manager Zhong Erxin has open-sourced iPhone Use, a toolkit that allows Codex to operate a real iPhone through natural-language commands. According to the project description, Codex can open apps, read screens, tap, swipe, type Chinese, search, and organize lists, while users watch the session live in the Codex sidebar. The stack is built on Appium’s WebDriverAgent, or WDA, which runs on Apple’s XCTest framework. Developers install WDA on the iPhone through Xcode, then use an MCP service on a Mac to relay Codex instructions to the device, where WDA carries out actions such as taps and text input. The setup does not require jailbreaking and does not need a separate Appium Server deployment. To cut the cost of repeatedly analyzing screenshots, iPhone Use first reads system-provided UI element data, including button text, position, and state. Screenshots are only sent back when controls cannot be identified or are blocked. The official guide still requires a USB connection, and the project has not yet released a full wireless setup.