Verified project record

callstack/agent-device

callstack/agent-device is a device automation CLI that gives AI coding agents structured accessibility snapshots, deterministic actions, and inspectable artifacts across mobile, TV, and desktop platforms. It acts as an execution and evidence layer rather than an autonomous decision-maker, requiring an external coding agent or QA harness to drive the interaction loop.

Project overview

It grounds AI coding agents in actual app feedback instead of relying on code reasoning, providing structured accessibility data, UI interaction capabilities, and evidence capture for iOS, Android, tvOS, Android TV, macOS, and Linux.

Project type: MCP · AI Agent
Use cases: Coding & Development · Automation
Deployment: Refer to project documentation
License: MIT

Best for

Developers and AI engineers who use AI coding agents and need to ground their agents in real device feedback.

Key capabilities

Inspects real app UI through structured accessibility snapshots, interactive refs like @e3, selectors, and React Native component trees.
Interacts by opening apps, tapping, typing, scrolling, performing gestures, waiting, asserting state, handling alerts, and closing sessions.
Captures evidence with screenshots, videos, logs, traces, network traffic, audio-level probes, performance samples, crash context, and React profiles.
Replays workflows by recording .ad scripts for local runs, CI, repeatable e2e checks, and strict Maestro YAML export.
Runs across platforms with iOS Simulator automation, Android Emulator automation, physical devices, tvOS, Android TV, macOS, Linux, and desktop app automation.
Node consumers can use the typed client and public subpaths for bridge integrations.

Limitations and risks

Minimal --platform web support reuses agent-browser.

Getting started

Installation requires medium difficulty setup, including coding or agent setup context and platform-specific SDKs such as Xcode or the Android SDK. Users need Node.js 22+ and can verify the setup using the doctor command.

Alternatives and comparisons

Provides an opinionated plugin and CLI for multi-agent orchestration with automated model routing, rather than a device automation CLI.
Offers a collection of open-source AI agents and RAG pipelines for development reference, rather than a tool for device UI verification.
Focuses on preserving AI agent context across sessions, rather than executing or verifying actions on local devices.

Project comparisons

Evidence and sources

README: Mobile app verification for AI agents. A device automation CLI for real apps on iOS, Android, TV, web, and desktop. Agents get token-efficient snapshots, semantic refs, and eviden…
GitHub project description: CLI to control iOS and Android devices for AI agents
README: - **Inspect** real app UI through structured accessibility snapshots, interactive refs like `@e3`, selectors, and React Native component trees. - **Interact** by opening apps, tap…
README: | Path | Best for | Start with | | --- | --- | --- | | Local | Exploration, debugging, and development loops on simulators, emulators, physical devices, macOS apps, and Linux desk…
README: Used by teams and developers at Callstack, JPMorgan Chase, Expensify, Shopify, Kindred, Total Wine & More, LegendList, HerLyfe, App & Flow, and more.

Reviewed at: 2026-07-16T21:55:27Z

Open original GitHub project