Your coding agent can now write AND run your E2E tests.
This is Claude driving the OpenTable Android app on an emulator, testing that the main booking flow works end to end.
AGENTIC UI TESTING that generates deterministic flows.
With Maestro MCP, your coding agent writes the flow. From there, it runs without AI, every CI run, every time you ship.
Asked Claude to find Austin rentals on Zillow's iOS app. With Maestro MCP, it opened the app, ran the search, scrolled through results. Zero taps from me.
Closing the agentic feedback loop on mobile separates agents that write code from agents that ship validated features.
Ship a feature. Your agent drives the device through Maestro MCP and validates it works. When a test step fails, it self-heals on the spot. The successful flow
let claude, cursor, codex, or any coding agent WRITE and RUN your UI tests using Maestro MCP, and WATCH every step live!
just ask: open the maestro viewer
Your coding agent now has a mobile device! 📱
Maestro CLI 2.6.0 ships Maestro Viewer: an iOS simulator, Android emulator, or physical device, live inside your coding agent.
Open the Maestro Viewer while your agent builds a mobile feature, or while it runs your Maestro tests. Watch every step and interact with the app live.
To open it, ask your coding agent:
> open the maestro viewer
Install Maestro MCP to get started:
Also in 2.6.0:
• Parallel local iOS simulator runs now work
• Cleaner relative paths across CLI and Cloud
• iOS XCTest logs in debug output dir
• maestro hierarchy outputs pure JSON
• HTML report links to Cloud + binary ID
Full release notes: