Skip to content
Stork.AIstork.ai
AI Tool

agent-device Review

agent-device is an open-source command-line interface (CLI) tool designed to enable AI agents to interact with real applications across iOS, Android, tvOS, macOS, and Linux platforms.

shipped May 15, 2026updated May 27, 2026aifreemium
ai
agent-device — product screenshot

Why it matters

1Version 0.15.0 released on June 26, 2026, introduced breaking changes for app discovery and subpath access.
2Supports automation across 6 distinct platforms: iOS, Android, tvOS, Android TV, macOS, and Linux.
3Provides token-efficient accessibility snapshots for AI agents to understand UI elements.
4Facilitates comprehensive debugging evidence collection, including screenshots, video, logs, and performance data.

Stork’s verdict on agent-device

agent-device gives AI agents structured control over diverse mobile and desktop apps, but its CLI nature demands technical proficiency and integration.

agent-device reviewed by Stork AI · stork.ai/en/agent-device

About agent-device

Business Model
Subscription SaaS
Platforms
Web
Target Audience
Developers and QA teams

overview

What is agent-device?

agent-device is an AI mobile testing tool developed by Callstack that enables AI agents, developers, and QA professionals to control iOS, Android, tvOS, macOS, and Linux desktop devices and apps programmatically. It automates app testing and UI interaction by taking accessibility snapshots, extracting UI elements, and executing taps, scrolls, and text input. The tool provides AI agents with 'eyes' on the UI through token-efficient snapshots of the accessibility tree, assigning stable references to elements, allowing agents to 'act with semantic precision' using these references or semantic selectors (e.g., by text, label, or role).

features

Key Features of agent-device

agent-device provides a robust set of features designed to give AI agents structured control and diagnostic capabilities over various application interfaces. These features enable precise interaction and comprehensive data collection for testing and automation workflows.

  • Structured control over mobile, TV, and desktop applications.
  • Token-efficient accessibility snapshots for UI understanding.
  • Semantic targeting of UI elements by text, label, or role.
  • Typed client access for programmatic interaction.
  • Comprehensive diagnostics, including screenshots, video, logs, and performance data.
  • Replayable workflows for automated regression testing.
  • Official agent-device skills for mobile automation.
  • UI element extraction and state verification.
  • Execution of taps, scrolls, and text input actions.

use cases

Who Should Use agent-device?

agent-device is primarily designed for AI agents, developers, and QA professionals who require programmatic control and automation capabilities over mobile, TV, and desktop applications. Its functionalities are tailored for advanced testing, debugging, and AI-driven interaction scenarios.

  • AI Agents: For structured control over mobile, TV, and desktop apps, enabling them to understand and interact with user interfaces.
  • Developers: For giving AI coding agents a real app feedback loop during feature implementation and for debugging regressions with comprehensive evidence.
  • QA Professionals: For AI mobile testing and QA of React Native and Expo apps, bug reproduction, regression testing, and clean-state testing across various platforms.

how to use

How to Use agent-device

agent-device operates as a command-line interface (CLI) tool, allowing users to install it and then issue commands to interact with connected devices and applications. It leverages accessibility trees to provide AI agents with structured information about the UI.

  • 1Install the agent-device CLI tool via npm or yarn.
  • 2Connect a target device (iOS, Android, tvOS, macOS, or Linux) to the development environment.
  • 3Use agent-device commands to take token-efficient snapshots of the application's accessibility tree.
  • 4Identify UI elements using semantic selectors or stable references provided by the snapshots.
  • 5Execute actions such as taps, scrolls, and text input on identified UI elements.
  • 6Collect debugging evidence like screenshots, video, logs, and performance data for analysis.

pricing

agent-device Pricing & Plans

The core agent-device CLI tool is an open-source library, making it free to use for developers and organizations. Callstack has announced upcoming 'AI mobile QA. Zero configuration' features, which are expected to be a paid offering. These 'Coming Soon' features aim to provide out-of-the-box autonomous testing workflows with pre-built templates for various infrastructures like GitHub Actions or EAS (for Expo builds), and a one-click agent setup, suggesting a subscription-based model for these advanced services.

  • Core CLI: Free (Open-source)
  • AI Mobile QA: Coming Soon (Early access available upon request, expected to be a paid subscription service)

Pros

  • +Provides structured, AI-native control over mobile, TV, and desktop applications.
  • +Offers broad cross-platform support: iOS, Android, tvOS, Android TV, macOS, and Linux.
  • +Leverages token-efficient accessibility snapshots for precise UI understanding by AI agents.
  • +Facilitates comprehensive debugging with evidence collection including screenshots, video, logs, and performance data.
  • +Supports advanced testing scenarios such as bug reproduction, regression testing, and clean-state testing.
  • +The core CLI is open-source, allowing for flexible integration and community contributions.

Cons

  • −Specific user reviews and reception are not widely available, making it challenging to assess real-world adoption and satisfaction.
  • −Advanced AI mobile QA features, such as 'Zero configuration' autonomous testing, are currently 'Coming Soon' and not fully released.
  • −Primarily a command-line interface, which may require technical proficiency and integration effort compared to GUI-based tools.
  • −Less established in the broader automation landscape compared to long-standing frameworks like Appium.
  • −Requires integration into existing development or CI/CD workflows for full automation benefits.

Similar Tools

agent-device vs Competitors

agent-device positions itself as a mobile verification layer specifically for AI agents, drawing parallels to browser automation patterns that enabled web agents. It aims to provide a structured, cross-platform CLI for interacting with real applications, differentiating itself from traditional mobile automation frameworks.

1
Appium (with AI extensions/integrations)↗

Appium is an open-source, highly extensible framework that supports a vast array of mobile, desktop, and TV platforms, now enhanced with AI for improved object recognition and self-healing capabilities.

Appium is an open-source framework, similar to agent-device's CLI nature, offering broad cross-platform support including iOS, Android, macOS, Windows, and TV, directly matching agent-device's listed platforms. Its AI capabilities are often added via extensions or integrations, whereas agent-device is described as 'Agent-native CLI for AI mobile testing'.

2
Maestro↗

Maestro offers a human-readable YAML-based syntax for defining UI and end-to-end tests across mobile and web, with AI assistance for test generation and a lightweight Studio IDE.

Maestro provides a lightweight, open-source framework with a CLI-first approach and AI assistance, similar to agent-device's 'Agent-native CLI'. It focuses on Android, iOS, and web, which is a strong overlap, though it doesn't explicitly list tvOS, Android TV, macOS, or Linux as direct testing targets in the same comprehensive way agent-device does.

3
Momentic↗

Momentic provides an AI-powered testing platform that allows for plain-English test creation and self-healing tests, supporting both cloud and local CLI execution.

Momentic offers AI-driven test automation with a 'start for free' model and supports local CLI execution, aligning with agent-device's freemium and CLI aspects. It covers web and mobile, and extends to Windows desktops, offering a good range of platforms, though not explicitly tvOS, Android TV, macOS, or Linux for testing.

4
Panto AI↗

Panto AI is an AI-native mobile QA platform that translates natural-language flows into deterministic tests, emphasizing real-device execution, self-healing, and root cause analysis.

Panto AI directly competes with its AI-native approach to mobile testing and a free plan, similar to agent-device's freemium model. It supports Appium and Maestro workflows, indicating CLI compatibility, and focuses on iOS and Android, which are core to agent-device, but doesn't explicitly list the broader desktop/TV OS support.

Visit agent-device↗

AI Reputation Report

Is agent-device yours?

ChatGPT, Perplexity, Gemini, Claude & Grok answer buyer questions about agent-device every day. See whether they name agent-device — or send buyers to a rival.

See what AI saysfree preview