Skip to content
All products
Screenwright

Screenwright

Beta

Give your AI agent hands on a real phone.

An MCP server that lets coding agents drive physical Android and iOS devices — tapping, typing, swiping, and actually seeing the screen.

Most mobile automation breaks the moment a layout shifts, because it depends on brittle selectors and hardcoded coordinates. Screenwright takes a different route: it exposes a real device as a set of Model Context Protocol tools, and pairs them with an on-device vision engine so an agent can look at the screen and decide what to do next. It runs against physical hardware over ADB and Xcode's device toolchain, not a simulator approximation.

Features

Real devices, not mocks

Drives physical phones and simulators over ADB and xcrun. What the agent sees is what a user would see.

Vision, not selectors

A bundled OCR and element-detection pipeline reads the screen directly, so automation survives layout changes.

Sixty-plus MCP tools

Tap, swipe, type, screenshot, read logs, set mock locations, inspect the view hierarchy — all exposed as callable tools.

Installs in one command

Published on npm and self-registers with your MCP client on install.

Related products

OrganiclawIn development

Organiclaw

A control plane for fleets of mobile bots.

Run many AI-driven bots across real devices from one dashboard — each one reads the screen and acts like a person would, with a full audit trail.

  • Web
  • CLI
  • MCP
Learn more
StoreShot EasyBeta

StoreShot Easy

Store screenshots that do not take a designer a week.

A browser-based editor for App Store and Play Store screenshots — device mockups, text, badges, backgrounds, exported to every size and language at once.

  • Web
  • MCP
Learn more
TerminaloIn development

Terminalo

Your terminals, files, and tabs in one menu-bar popup.

A macOS menu-bar workspace that keeps terminals, file browsers, and web views alive side by side — and lets an AI agent drive all of them.

  • macOS
  • MCP
Learn more