Handsets: High-Performance Android Control CLI for Agents

What changed

The Handsets CLI, developed in Rust, provides a high-performance interface for driving Android devices directly from the shell. It aims for millisecond-level command execution, a substantial improvement over traditional methods like adb shell which can take hundreds of milliseconds. Handsets achieves this by using a custom wire protocol over adb forward, employing length-prefixed binary frames for communication between the host and the device daemon. The daemon itself is a single, small JAR file that runs on the device without requiring root access or a separate app installation.

Key features include:

• Low Latency: Single-call latency is reported between 2-7 ms, compared to 40-700 ms for adb shell and 30-100 ms for uiautomator2.
• On-Device Daemon: A minimal, single-jar (~few hundred KB) daemon simplifies deployment.
• Flexible Control: Commands can be executed from any language by calling the hs binary as a subprocess, with JSON output available for programmatic parsing. Python bindings are also provided via pip install handsets.
• Advanced Selectors: A CSS-like selector syntax, inspired by Playwright, allows for precise targeting of UI elements based on attributes, text content, and relative positioning.
• hs tui: A new terminal user interface that streams the device's interactive elements in real-time. This allows users to navigate and interact with the device using keyboard shortcuts, with actions mirroring the scripting commands for easy transition from interactive use to automation.
• Efficient Imaging: Supports JPEG screenshots (default 768px long-edge) and WebP for compact lossy exports, with PNG reserved for debug purposes.

Handsets differentiates itself from tools like uiautomator2 and Appium by focusing on lean CLI operation and minimal per-call overhead, eschewing features like recorders, HTML reports, or iOS support in favor of speed and simplicity for tap-heavy automation.

Why it matters for builders

For AI agents and developers building them, Handsets offers a critical reduction in the feedback loop for Android automation. The millisecond-level latency means agents can react to screen changes and execute commands much faster, enabling more complex and responsive interactions. This efficiency is crucial for tasks requiring rapid decision-making or precise timing, such as UI testing, automated workflows, or even game automation.

Practical impact

Builders can integrate Handsets into their automation pipelines by installing it via a curl script or using the Python package. The CLI's hs use command initiates the on-device daemon, after which commands like hs tap "Continue" or hs wait "Welcome" can be executed. For more complex interactions, the hs ui command provides a table of tappable nodes, which can be piped into AI models to determine the next action. The new hs tui command offers an interactive way to explore and script flows, with the displayed commands directly usable in automation scripts. Developers can experiment with the CSS-like selectors for precise element targeting and leverage the JSON output for robust scripting in various programming languages.

Caveats and source limits

The project is described as a "fresh release" and is pre-1.0, indicating potential for API changes. While benchmarks are provided comparing Handsets to adb shell, uiautomator2, and Appium, these are self-reported and not independently verified. The source does not specify pricing, as it is an open-source project. Support for platforms beyond macOS and Linux is not detailed. The provided metadata indicates 67 stars and 4 forks on GitHub, suggesting a nascent but growing community interest.