Loadmill Droid CUA

Loadmill's Droid CUA (Computer-Using Agent) is a desktop application and CLI tool that lets mobile engineering teams create, run, and manage tests for Android devices and iOS simulators using plain-English instructions instead of traditional automation code. It is built by Loadmill, a company focused on generative AI for test automation that unifies API, UI, and performance testing. The desktop app is currently available for macOS Apple Silicon, with the latest release at v2.22.0 as of June 2026.

What It Is

Droid CUA is a computer-using agent for mobile testing. Rather than requiring engineers to write selector-based scripts or maintain brittle UI hierarchies, the agent accepts natural language descriptions of test scenarios—such as "test signup with valid and invalid email flows"—and autonomously navigates the app, performs actions, and generates reusable test scripts. Those scripts can be reviewed, edited, saved, and rerun from the desktop app or executed headlessly in CI pipelines via the CLI.

How the Agent Works

The agent follows a structured execution loop:

• Connects to an Android device or emulator via ADB, or to an iOS simulator via Appium and XCUITest on macOS
• Captures full-screen device screenshots and scales them down for model compatibility
• Sends screenshots and user instructions to an AI model (requires an OpenAI API key)
• Receives structured actions—click, scroll, type, keypress, wait, drag—and rescales outputs to real device coordinates
• Executes actions on the device, validates assertions, handles failures, and repeats until task completion
• Saves results as JUnit XML output for CI systems and maintains a project run history inside the desktop app

Desktop-First Workflow with CI-Ready Output

The primary interface is a macOS desktop application that guides setup, manages device connections, and provides live execution logs during test runs. From the same app, teams can review past run reports stored in a project results folder. For CI, scripting, or advanced workflows, the CLI (@loadmill/droid-cua on npm) supports headless execution of saved instruction files with options for device targeting, platform selection, recording, and debug output.

Platform and Setup Requirements

• Android: Physical devices and emulators via ADB; Android Emulator CLI for launchable emulators
• iOS: Simulators on macOS only, requiring Xcode with iOS Simulator, Appium, and the XCUITest driver
• All platforms: An OpenAI API key or a Loadmill account login

The desktop app is currently distributed as a macOS Apple Silicon DMG via GitHub Releases. The CLI is available via npm (npm install -g @loadmill/droid-cua).

Update: v2.22.0

The latest release is droid-cua v2.22.0, published June 29, 2026. The repository was created in March 2026 and has seen active release cadence since then, indicating the product is in active early development. The homepage also highlights a new AI Mobile Testing Agent as a recently launched capability, consistent with the GitHub release timeline.

Why It Differs from Traditional Mobile Automation

Traditional selector-based mobile automation breaks when UI elements change, requiring ongoing maintenance that consumes engineering time. Droid CUA operates through visible screen context rather than UI hierarchies or locators, which the product page describes as making tests less fragile across releases. The agent starts from intent—a Jira story, epic, or plain-English requirement—and produces reusable test assets without requiring automation code to be written manually.