iphone-use
An open-source iOS automation framework that allows AI agents to directly control physical iPhone devices via WebDriverAgent without requiring application APIs. It supports MCP Server integration and is released under the MIT license.
Timeline (last 90 days)
iphone-use is an open-source project that allows AI agents to directly control a real iPhone without requiring an API.
The core idea of iphone-use is to convert the iPhone screen into textual information for the model to understand, and then execute real actions such as tapping, typing, and scrolling via WebDriverAgent.
iphone-use provides three-state feedback for each action: applied, not sent (sent but not effective), and unknown.
To address the issue of payment apps blocking screenshots, iphone-use returns interface structure information in wireframe form.
iphone-use supports task recording and playback; once a task is successfully executed once, it can be solidified into a repeatable flow that no longer requires calling the model during playback.
iphone-use offers three integration methods: an HTTP API, a 21-tool MCP Server (designed specifically for Claude Code), and remote control (via a browser or iOS app).
iphone-use is open-sourced under the MIT license.
iphone-use received 70 votes on Product Hunt and ranked 13th on the day's leaderboard.
The accuracy of screen-to-text understanding and the information loss in wireframe mode may affect the success rate of complex interface tasks.
All Facts (9)
iphone-use is an open-source project that allows AI agents to directly control a real iPhone without requiring an API.
50%UnverifiedThe core idea of iphone-use is to convert the iPhone screen into textual information for the model to understand, and then execute real actions such as tapping, typing, and scrolling via WebDriverAgent.
50%Unverifiediphone-use provides three-state feedback for each action: applied, not sent (sent but not effective), and unknown.
50%UnverifiedTo address the issue of payment apps blocking screenshots, iphone-use returns interface structure information in wireframe form.
50%Unverifiediphone-use supports task recording and playback; once a task is successfully executed once, it can be solidified into a repeatable flow that no longer requires calling the model during playback.
50%Unverifiediphone-use offers three integration methods: an HTTP API, a 21-tool MCP Server (designed specifically for Claude Code), and remote control (via a browser or iOS app).
50%Unverifiediphone-use is open-sourced under the MIT license.
50%Unverifiediphone-use received 70 votes on Product Hunt and ranked 13th on the day's leaderboard.
50%UnverifiedThe accuracy of screen-to-text understanding and the information loss in wireframe mode may affect the success rate of complex interface tasks.
50%