iphone-use: open-source tool lets an AI agent drive a real iPhone

The MIT-licensed iphone-use project lets an AI agent control a real iPhone via Apple's WebDriverAgent: it reads the screen as text, taps, swipes and types. On an iPhone 17 Pro Max tap-and-settle dropped from 8.5 s to 4.2 s and cold connect from 75 s to 13 s.

DEV Community

Open source

A Mac daemon runs WebDriverAgent on a USB-connected iPhone, exposing an HTTP API and an MCP server with 21 tools
Every action reports whether it was applied, not sent or unknown, and whether a retry is safe
Protected payment-app screens are returned as a labelled accessibility-tree wireframe
A recorded task can be replayed in one command with no model in the loop

#Apple#iPhone#Claude

OpenAI lets ChatGPT subscribers use their plan in third-party coding tools

Caveman: open-source tool cuts AI coding agent output by 65%

Equals Money lets customer AI tools read data but not move money

Hardware

iphone-use: open-source tool lets an AI agent drive a real iPhone

OpenAI lets ChatGPT subscribers use their plan in third-party coding tools

Caveman: open-source tool cuts AI coding agent output by 65%

Equals Money lets customer AI tools read data but not move money

Tier IV open-sources self-driving AI chip design to challenge Nvidia