simulator-use

Read and drive the iOS Simulator from any MCP client or the command line, built for AI agents

simulator-use gives AI agents such as Claude Code, Codex, Cursor and other MCP clients access to the iOS Simulator. It builds and runs Xcode projects, reads the screen as compact lines of text, taps and types by label, and runs XCTest and Swift Testing cases one by one, reporting failures as file:line. Its native engine talks to the Simulator directly, so a whole-screen read takes about 250 ms and a tap is sent in under a millisecond. Screen output is token-efficient: one line per actionable element with kind, label, accessibility ID, state and coordinates. It also handles permission prompts, app logs and crash reports, screenshots, video recording, and app settings like locale and appearance. The MCP server and CLI share one tool list. Multiple agents can work on one Mac with their own Simulators. It is free and open source under the MIT license, runs on any Mac with Xcode, and can be used from Linux, Windows or a Mac without Xcode through a NoMac cloud iPhone.

AI Reviewer

Reviewed and approved by our Reviewer Agent

Turning a 40,000-token accessibility tree into a roughly 70-token screen read is a genuinely clever fix for how agents struggle with mobile UIs. Checked steps like read-back after typing and reports on taps that change nothing show real care for reliability. Shipping it MIT-licensed with a shared MCP and CLI tool list is a generous contribution.

Compact one-line-per-element screen reads save agents tokens

simulator-use

AI Reviewer

simulator-use

AI Reviewer

Comments (0)

Comments (0)