Google Expands Android Developer Tooling with CLI Device Streaming and New AI Agent Skills

As the landscape of mobile development continues to evolve with the integration of artificial intelligence, Google is giving Android developers greater flexibility in how they build applications. Recognizing that developers rely on a diverse mix of agents, large language models, specialized tools, and command-line interfaces, the company has announced a major update to its development ecosystem. This release focuses on bridging the gap between modern AI-driven workflows and the rigorous demands of native platform development, ensuring that developers can build high-quality applications regardless of their preferred toolchain.

The centerpiece of this announcement includes significant updates to the Android CLI command-line tooling, most notably the integration of Android Device Streaming. Furthermore, Google is expanding its repository of Android skills to help AI agents better understand platform-specific standards, featuring a deep dive into the Wear OS Compose Material 3 skill. By providing these resources, Google aims to streamline the development process and empower AI agents to handle complex engineering tasks with greater accuracy and efficiency.

Android CLI for Command-Line Development

The Android CLI serves as a foundational tool designed to make command-line interface development significantly easier for mobile engineers. Built to be agnostic and flexible, it supports any AI agent or third-party tool looking to build applications for the Android platform more efficiently.

By leveraging the android-cli skill alongside the underlying command-line tooling, AI agents can autonomously execute a wide array of project management tasks. These capabilities extend from creating, building, testing, and managing complex Android projects to helping developers set up their local development environments. Additionally, the CLI allows agents to create and run emulators as well as execute automated test runs directly from the terminal, minimizing friction and keeping developers in their preferred command-line workflows.

Android Device Streaming Now Available in CLI

One of the persistent challenges in mobile application development is ensuring that software performs reliably across a fragmented hardware and operating system landscape. While emulators are invaluable for rapid iteration, testing on real physical devices remains critical for catching hardware- or OS-specific bugs that might otherwise slip through to production. However, maintaining an extensive physical device lab locally is often impractical or outright impossible for many developers.

To address this challenge, Google is bringing Android Device Streaming directly into the Android CLI, allowing developers and their AI agents to access real physical devices remotely. This integration means that an AI agent can now utilize Android Device Streaming from anywhere within the terminal environment.

Through a secure Android Debug Bridge over SSL connection, an agent can interact with physical devices just as if they were physically plugged into the workstation via a USB cable. This powerful capability allows agents to spin up remote devices, deploy builds, collect diagnostic logs and traces, and even capture screenshots headlessly—all without leaving the command line. To begin using this feature, developers simply need to link their project, instruct their agent to list available remote devices, and select the target hardware required for their specific testing scenario.

Grounding Agents with Android Skills

While large language models possess vast general knowledge, they frequently struggle with platform-specific standards, rapidly changing APIs, and the latest architectural patterns. To bridge this gap between general model training and official platform guidance, Google continues to expand its comprehensive Android skills repository.

Android skills consist of structured instructions, organized as specific markdown files, designed to ground AI agents using official documentation directly from the developer portal. Instead of relying solely on a model’s static training cutoff date, these skills inject precise, up-to-date data, official API references, code samples, and modern architectural patterns directly into the agent’s active context window.

With more than twenty distinct skills currently available in the repository, developers can equip their AI agents to navigate increasingly complex and specialized development tasks with confidence. The engineering behind these skills is rigorous, backed by thorough evaluations to ensure reliability. Furthermore, Android skills are designed to be entirely environment-agnostic. Whether a developer is writing code inside Android Studio, utilizing specialized environments like Antigravity, or pairing with third-party coding agents such as Claude and Codex, the Android skills integrate seamlessly across the entire development setup. Managing these skills across various projects and agent directories is handled smoothly through the Android CLI.

Skill Spotlight: Wear Compose Material 3

Developing applications for Wear OS presents unique design and engineering challenges compared to traditional smartphone development. Developers must account for round viewports, rotary input mechanisms, ambient display modes, strict power consumption limits, and specialized UI components such as the TransformingLazyColumn, AppScaffold, and ScreenScaffolds containers.

Without explicit, platform-specific guidance, standard large language models frequently lack the contextual understanding required to implement these distinct patterns correctly, often missing the nuances that make smartwatch applications functional and polished. To solve this, Google introduced the Wear Compose Material 3 skill, specifically engineered to guide agents through the intricacies of building modern Wear OS interfaces.

Early adopters across the industry are already reporting substantial productivity gains after integrating this skill into their workflows. For instance, the engineering team at FotMob utilized the Wear Compose Material 3 skill to modernize their existing application built on earlier Material 3 iterations. Their migration tasks included updating multiple lists to utilize TransformingLazyColumn paired with ScreenScaffold content padding, implementing ListHeader titles, applying SurfaceTransformation effects to cards and buttons, refining theme typography, and generating accurate Wear previews.

The results of this integration were immediate and impactful. The resulting code changes compiled successfully and were thoroughly verified on emulators for smooth scrolling, rotary input behavior, edge morphing, and right-to-left language support. This allowed the engineering team to successfully remove legacy wrappers alongside a significant amount of custom rotary and focus boilerplate code.

Notably, the skill successfully caught subtle implementation mistakes that the underlying AI model had missed entirely, such as forgetting to forward the contentPadding from the ScreenScaffold into the underlying list, and correctly prioritizing theme typography over hardcoded density-independent pixels. Reflecting on the efficiency gained during the process, Roy Solberg, an Android Tech Lead at FotMob, noted that a single skill utilized over the course of an afternoon successfully facilitated the migration of eight separate lists while eliminating a pile of custom rotary code.