iPhone Use

•
480
•

Control and automate physical iPhones and iOS simulators through AI agents or remote browser interfaces using accessibility tree navigation and native touch inputs.

iPhone Use allows AI agents—such as Claude Code and Codex—to interact directly with physical iOS devices and simulators without requiring official app APIs. Powered by a self-hosted Rust daemon on macOS and a lightweight XCTest runner on the target device, it translates the iPhone display into an accessibility text tree rather than parsing raw pixels, significantly reducing token consumption and execution latency. It supports touch gestures, text entry, and app navigation over USB or Wi-Fi, while providing companion browser and iOS apps for human remote control, live H.264 screen viewing, and zero-token replayable testing flows.

主要功能

01Reads device screens as accessibility text trees to minimize token consumption and provide 0.1s inspect speeds.
02Performs native taps, swipes, and bulk text input up to 20,000 characters over USB or encrypted Wi-Fi.
03Automates any iOS application, including health, banking, and apps that block standard screen captures.
04Provides low-latency H.264 live streaming for remote human control via web browser or a paired iOS app.
05429 GitHub stars
06Generates zero-token, replayable automation flows from recorded actions for fast, deterministic execution.

使用案例

01Automating end-to-end testing and QA regression flows on real iOS hardware or iOS Simulators.
02Enabling AI coding assistants to extract information or perform workflows inside native mobile apps lacking public APIs.
03Remotely operating a physical iPhone from a desktop browser or another iOS device over local networks and secure tunnels.