Google Announces Major Android CLI and AI Agent Tooling Updates with Remote Real-Device Streaming and Specialized Skills

As the landscape of mobile app development continues to evolve rapidly, Android developers are increasingly turning to a diverse array of agents, large language models (LLMs), tools, and command-line interfaces (CLI) to streamline their workflows. Recognizing this shift, Google has announced a suite of major updates designed to empower developers regardless of how they choose to build, with a primary focus on enhancing command-line tooling, expanding agent capabilities, and introducing specialized knowledge repositories for modern Android development.

The latest announcements center around significant upgrades to the Android CLI command-line tooling—including the integration of Android Device Streaming—alongside an expanding library of structured Android skills aimed at grounding artificial intelligence agents in official platform best practices. By bridging the gap between default model training data and current platform standards, these tools aim to help developers build beautiful, high-quality Android applications more efficiently than ever before.

Android CLI for Command-Line Development

At the heart of the latest tooling updates is the Android CLI, designed specifically to make command-line interface development smoother and more intuitive for modern software engineers. The utility is engineered to support virtually any AI agent or external tool in building efficiently for the Android ecosystem.

When paired with the designated android-cli skill, AI agents can leverage the Android CLI to autonomously handle a comprehensive range of development tasks. This includes creating, building, testing, and managing complex Android projects, assisting with the initial setup of development environments, creating and running emulators, and executing automated test runs directly from the terminal. By centralizing these capabilities into a unified command-line experience, developers can integrate AI assistants seamlessly into their existing command-line workflows without needing to constantly switch contexts between graphical user interfaces and terminal windows.

Android Device Streaming Now Available in CLI

One of the most persistent challenges in mobile software engineering is ensuring that applications perform reliably across a fragmented hardware and operating system landscape. While testing on real physical devices remains critical for catching hardware- and OS-specific issues, developers do not always have physical access to the exact device variants they need in their local development environments.

To address this friction, Google is bringing Android Device Streaming directly into the Android CLI, allowing developers and their AI agents to access real physical devices remotely. Through a secure Android Debug Bridge (ADB) over SSL connection, an AI agent can now interact with physical hardware as if the device were physically plugged into the developer’s workstation via a USB cable.

This powerful integration enables developers and automated agents to spin up remote devices, deploy builds, collect logs and traces, and even capture screenshots headlessly—all executed seamlessly through the terminal interface. To get started with the feature, developers simply need to link their project, instruct their agent to list the available remote devices, and select the specific target hardware they require for their testing cycle. This capability significantly reduces the overhead of maintaining physical device farms locally while ensuring that both human engineers and AI assistants can verify builds against actual hardware configurations.

Grounding Agents with Android Skills

As the capabilities of large language models expand, a recurring hurdle in software development is the gap between a model’s default training data and the rapid, continuous updates of platform-specific standards. To combat hallucinations and outdated code patterns, Google continues to grow its comprehensive Android skills repository, designed to inject precise, up-to-date knowledge directly into an agent’s working context.

Android skills are structured instructions, formatted as specific SKILL.md files, that ground AI agents using official guidance sourced directly from developer.android.com. Rather than relying solely on a model’s training cutoff, these structured files provide fresher data, accurate API references, relevant code samples, and modern architectural patterns. With more than 20 distinct skills currently available in the public repository, developers can equip their AI agents to handle increasingly complex and specialized development tasks with high fidelity.

The philosophy and methodology behind this initiative are rooted in rigorous evaluation. Google emphasizes that robust platform tooling must be accompanied by thorough evaluations to ensure reliability, an approach detailed in their documentation regarding building tools designed to handle platform deprecations and rapid iterations.

Managing these skills across different projects and individual agent directories is streamlined directly through the Android CLI. Furthermore, these skills are built to be completely environment-agnostic. Whether a developer is writing code inside Android Studio, utilizing Antigravity, or pairing with third-party coding agents such as Claude and Codex, the official Android skills function consistently across the entire developer setup.

Skill Spotlight: Wear Compose Material 3

Building applications for Wear OS presents unique engineering and design considerations that differ significantly from standard mobile development. Developers must account for round viewports, rotary input mechanisms, ambient display modes, strict power consumption limits, specialized layout containers like AppScaffold and ScreenScaffolds, and performance-optimized components such as TransformingLazyColumn.

Without explicit, platform-specific guidance, standard LLMs frequently lack the nuanced understanding required to implement these distinct patterns correctly, often missing the details that make Wear OS applications polished and performant. To bridge this knowledge gap, Google released the specialized Wear Compose Material 3 skill (wear/wear-compose-m3).

Early adopters of the Wear Compose Material 3 skill are already reporting substantial productivity gains in real-world production environments. For instance, the engineering team at FotMob utilized the skill to tackle extensive modernization tasks on their existing Wear M3-based application. Their migration efforts included updating multiple lists to use TransformingLazyColumn with proper ScreenScaffold content padding, implementing ListHeader titles, applying SurfaceTransformation across cards and buttons, adopting modern theme typography, and configuring accurate Wear previews.

The results of integrating the specialized skill into their workflow were immediate and impactful. The resulting code changes compiled successfully on the first pass and were thoroughly verified on the emulator for smooth scrolling, rotary input behavior, edge morphing, and right-to-left (RTL) layout support. Crucially, the process allowed the team to safely remove their legacy wrappers alongside a significant accumulation of custom rotary and focus boilerplate code.

Additionally, the skill successfully caught subtle implementation mistakes that the underlying model would have otherwise missed, such as forgetting to forward the ScreenScaffold content padding directly into the list components, and correctly prioritizing theme typography over hardcoded sp values. Highlighting the efficiency of the workflow, Roy Solberg, Android Tech Lead at FotMob, noted the dramatic scope of the update achieved in a remarkably short timeframe, emphasizing the value of targeted, platform-grounded AI assistance in modern app development.