Phone Harness

Phone Harness connects an AI agent to a real iPhone or Android so it can use apps through their screens. It's for developers who want Claude Code, Codex, or another agent to carry out phone tasks without a dedicated app API. The software is open source under the MIT license, and it doesn't require a jailbreak or an app installed on the phone.

With your own device, the harness runs on your computer. iPhone control uses the Mac's iPhone Mirroring window and can work in the background. Android control works from macOS, Linux or Windows through adb, over USB or Wi-Fi, without a mirrored window. You bring your own agent; the harness supplies phone interaction rather than an AI model.

The agent can tap, type, swipe and inspect the result. On iPhone, Apple's Vision framework reads screen text; on Android, the phone's accessibility tree supplies it. Both platforms use the same control helpers. For unlabeled icons and controls, the agent needs a screenshot and a vision-capable model because text recognition alone won't identify them.

You can connect your own phone or rent a hosted iPhone or Android through Phone Harness Cloud. Rented devices run remotely; control of your own device uses the connected phone and computer.

Some tasks still need you. Unlocking an iPhone pauses mirroring, and a PIN-locked Android needs user input. The harness doesn't support multi-touch, camera or Face ID flows, and DRM-protected video appears black.