Skip to content
simulator-use

simulator-useMCP Server

Released
v0.18.1
MIT

Summary

MCP server that lets coding agents build, run, read and drive the iOS Simulator: compact screen reads, tap-and-type steps, XCTest runs, logs and video.

Features

  • Builds, installs and launches an Xcode project in one call, with errors as file:line
  • Compact one-line-per-element screen reads instead of the full accessibility tree
  • Checked tap, fill, scroll, swipe and wait steps that stop at the first failure
  • Runs XCTest and Swift Testing per case and reports each failure
  • Logs, crashes, screen recording, permission prompts and StoreKit testing
  • Optional cloud iPhone Simulator for machines without Xcode

Installation

Set up this MCP server in your favorite AI agent — copy a ready-made configuration below.

Any MCP-compatible agent

Most agents (Claude, Cursor, Windsurf, VS Code, and more) read a standard mcpServers configuration.

  1. Open your agent's MCP configuration file.
  2. Merge the snippet below into it.
  3. Restart the agent — the "simulator-use" tools will be available.
{
  "mcpServers": {
    "simulator-use": {
      "command": "npx",
      "args": [
        "-y",
        "simulator-use",
        "mcp"
      ]
    }
  }
}

Description

simulator-use gives Claude Code, Codex, Cursor and other MCP clients control of the iOS Simulator. It builds and runs your Xcode project, reads the screen as a few lines of text, taps and types by label or accessibility identifier, and runs your tests case by case. It keeps its own native connection to the Simulator rather than starting a process for every call.

How the agent sees the screen

ui_look returns one line per actionable element (kind, label, #id, state and tap point in points), folding away groups, status bar and keyboard keys, and reports alerts and permission prompts in front of the app. Web views in UIKit apps are read too; unreadable areas come back as one line so the agent knows to take a screenshot. ui_act runs a list of steps (tap #signIn, fill #email …, scroll down, wait "Welcome") and checks each one, reading typed text back and stopping at the first failure with the current screen.

Tools

app_run (find project, build, install, launch; errors as file:line), devices, device_use, app_control (install, launch, permissions, location, push, appearance, locale, StoreKit testing), ui_look, ui_act, app_logs (print output, system log, crashes), tests_run (XCTest and Swift Testing, per case), screen_record, job_status, job_cancel, and cloud. The same tools exist as a CLI.

Requirements

macOS with Xcode, a Simulator runtime and Node 20+. Works with SwiftUI, UIKit and React Native apps; Expo, React Native, Flutter and Capacitor builds are made with their own tools and then installed. On Linux, Windows or a Mac without Xcode, the cloud tool runs the same tools on a NoMac cloud Mac, billed per minute; the local tools send nothing anywhere. MIT licence.

MCP: Argent

by Software Mansion

New

Software Mansion's MCP server for controlling, debugging and profiling iOS simulators, Android emulators and devices, TVs, Electron and native Mac apps.

MCP: agent-device

by Callstack

Featured

Give a coding agent a live feedback loop on real mobile, TV and desktop apps — inspect, tap, type and capture evidence on iOS, Android and HarmonyOS over MCP or a CLI.

MCP: Mobile MCP

by Mobile Next

Featured

Platform-agnostic MCP server that drives iOS and Android apps on simulators, emulators and real devices through the accessibility tree rather than screenshots.

Browse all MCP servers →