Open-Source iPhone Use Lets OpenAI's Codex Control a Physical iPhone
A new open-source Codex plugin bridges OpenAI's coding agent to a physical iPhone over USB, letting it drive real apps like WeChat, Taobao, and Meituan.
- iPhone Use connects OpenAI Codex to a real iPhone over USB via an MCP plugin
- Built on Appium's WebDriverAgent, pinned to commit 16.14.0, running fully on local loopback
- Exposes 17 tools: taps, swipes, Unicode input, batched actions, scroll-find, list collection
- Falls back to screenshot plus coordinate tapping when accessibility targeting fails
- Requires macOS, full Xcode, Apple developer account, Python 3.9+, Node 20.19+
- MIT licensed, 870+ stars; demos include WeChat, Taobao, Meituan, Xiaohongshu automation
iPhone Use lets Codex control a physical iPhone over USB
A developer has released iPhone Use, an open-source package that connects OpenAI’s Codex agent to a physical iPhone through USB. The repository passed 870 GitHub stars shortly after launch, while the author’s demo showed Codex scrolling WeChat Moments, navigating Taobao ordering flows, and summarizing group chats on the device.
The project targets mobile workflows that lack practical web equivalents, including Chinese super-apps such as WeChat, Taobao, Meituan, and Xiaohongshu. It gives Codex access to those apps through Apple’s existing UI-testing infrastructure, with no jailbreak or cloud-hosted device required.
WDA provides the hands
WebDriverAgent, commonly called WDA, executes taps, swipes, button presses, and text input through Apple’s XCTest framework. Appium has long used WDA for iOS automation; iPhone Use packages it behind a Model Context Protocol server so Codex can invoke device actions as tools.
The bundled Python service communicates with WDA through local loopback forwarding on port 18100. Remote addresses are rejected, and runtime data, downloaded dependencies, signing credentials, and builds remain on the Mac according to the project documentation.
Seventeen tools cover the control loop
The MCP server exposes 17 tools to Codex, with two additional tools reserved for the live screen widget. Representative tools cover observation, interaction, text entry, efficient multi-step work, and environment checks.
| Purpose | Tools | What they provide |
|---|
This story is for Pro members
You've reached the end of the free preview. Upgrade to AlphaSignal Pro to read the full article - and everything else behind the paywall.
Comments
