# callstack/agent-device

Mobile app automation and verification for AI coding agents. CLI, MCP server, and typed Node.js API for iOS, Android, HarmonyOS, TV, web, macOS, and Linux.

Repository: https://github.com/callstack/agent-device
Canonical: https://ross.abutalabs.com/products/agent-device
Homepage: https://agent-device.dev
Language: TypeScript
License: MIT
License Family: permissive
Topics: agentic-ai, agents, automation, mobile, testing, adb, android-emulator, e2e-testing, expo, ios-simulator, mcp, react-native, xcuitest, ai-agents, flutter, mobile-testing, performance-optimization
Last push: 2026-08-26T18:57:40+00:00

## Health v2 (maintenance only)
Score: 78/100 (v2, computed 2026-09-03T02:39:23.370411+00:00)
- activity 99, release rhythm 87, longevity 15
- inputs: {"age_days": 215, "days_push": 7, "days_rel": 9, "gap_med": 0, "n_releases_24m": 118}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4233, forks 265 (observed 2026-08-28T04:08:39.416808+00:00)

## What it is
A CLI, MCP server, and typed Node.js API that lets AI coding agents automate, inspect, and verify real mobile and desktop apps. It provides token-efficient accessibility snapshots, semantic element refs, and evidence capture across iOS, Android, HarmonyOS, TV, web, macOS, and Linux.

## Use cases
- let a coding agent verify its UI changes in a running mobile app
- automate iOS simulator and Android emulator testing from the terminal
- give AI agents eyes on a native app via accessibility snapshots
- run end-to-end flows on real devices and replay them as scripts
- coordinate device access across parallel agent worktrees
- capture bug evidence and profiling data from mobile apps
- connect agents to remote device clouds over MCP

## When to choose
- you want AI coding agents to test and verify native mobile apps automatically
- you need one automation interface across iOS, Android, TV, and desktop targets
- you prefer token-efficient accessibility-tree snapshots over screenshot-based reasoning
- you use Claude Code, Cursor, Codex, or another MCP-capable agent

## When to avoid
- you only need browser automation for web apps
- you need a traditional CI-only test framework without agent integration
- your stack requires Node.js older than 22.12

## Facets
- artifact type: cli-tool
- maturity: active
- function: testing, e2e-testing, cli, mcp, agent-framework, developer-tools
- domain: mobile-development, testing, developer-tools
- platform: cli, cross-platform
- tags: mobile-automation, mcp-server, ai-coding-agents, ios-simulator, android-emulator, react-native, flutter, accessibility-snapshot, device-automation, expo, automation, ai-agents, nodejs, macos, linux, ios, android, web-server

## Member repositories
- callstack/agent-device (main) score 78

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:39.416808+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:22:19.495209+00:00, confidence not recorded.
  - readme: https://github.com/callstack/agent-device (fetched 2026-08-28T04:08:39.416808+00:00, sha 13774164ad9e)
  - homepage: https://agent-device.dev (fetched 2026-08-29T09:12:27.944068+00:00, sha 1eddfa93453f)
- Data as of 2026-08-30T08:39:29.467469+00:00.