callstack / agent-device

A token-efficient device automation CLI that gives AI agents hands, eyes, and evidence collection capabilities across mobile, TV, and desktop platforms

活跃维护 MIT TypeScript Tracked
4.3k 266 15 小时前
CINPM
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

星标趋势

数据积累中,暂无足够数据生成趋势图

AI 分析

项目摘要

agent-device is a device automation CLI that enables AI agents to control and inspect real apps across iOS, Android, TV, web, and desktop platforms. It provides token-efficient accessibility snapshots, semantic element references, and rich evidence capture (screenshots, videos, logs, traces) to ground AI agent workflows in actual device behavior rather than code reasoning alone.

为什么值得关注

It bridges the gap between AI agents and physical device interaction, offering a purpose-built tool for agent-driven mobile testing with thoughtful design around token efficiency, structured accessibility data, and MCP integration. With 3,400+ stars and 100 releases in 6 months, it's rapidly becoming a key infrastructure piece for AI-agent-based app verification.

优势

  • Cross-platform support spanning iOS, Android, TV, web, and desktop with native, Expo, Flutter, and React Native compatibility
  • Token-efficient design optimized for AI agent consumption with semantic refs and structured accessibility snapshots
  • Comprehensive evidence capture including screenshots, videos, network traces, performance samples, and crash context
  • Strong engineering practices with CI, tests, 35 contributors, and active release cadence
  • MCP server support for seamless integration with AI agent ecosystems

局限性

  • Relatively new project (created January 2026) with a rapid release cycle that may indicate API instability
  • Web platform support is minimal compared to mobile, delegating to agent-browser for browser sessions
  • Requires physical devices, simulators, or emulators, adding infrastructure complexity for CI pipelines

使用场景

  • AI agent-driven end-to-end mobile app testing and verification on real devices
  • Automated debugging workflows where agents inspect UI state and capture diagnostic evidence
  • Cross-platform app verification with repeatable .ad script replay for CI pipelines
  • QA harness integration for autonomous mobile app quality assurance
目标用户: Mobile developers, QA engineers, and AI agent developers who need to verify app behavior on real devices through agent-driven workflows
学习曲线:
分析模型:LongCat-2.0 | 分析时间:1 个月前