Google Announces Major Android CLI Updates and Expands AI Agent Capabilities with Real Device Streaming and New Skills

As the landscape of mobile development continues to evolve with the integration of artificial intelligence, Android developers are increasingly turning to a diverse array of agents, large language models, tools, and command-line interfaces to build applications. Recognizing this shift, Google has announced a significant set of updates to its Android CLI command-line tooling. The primary goal of these enhancements is to empower developers to build beautiful, high-quality Android apps regardless of their preferred coding environment or AI assistant. Alongside these toolchain improvements, the company is introducing deeper integration for remote hardware testing and expanding its repository of Android skills, highlighted by a comprehensive deep dive into the Wear OS Compose Material 3 skill.

The core of this recent push centers on making command-line interface development more robust and accessible. The Android CLI is designed to serve as a versatile utility that supports any AI agent or external tool in building applications more efficiently. By utilizing the integrated android-cli skill, developers can instruct their preferred agents to execute a wide variety of project management tasks. These capabilities include creating, building, testing, and managing complex Android projects from the terminal. Furthermore, the tool assists developers in setting up their local development environments, spinning up and running virtual emulators, and executing automated test suites without needing to constantly context-switch to a graphical user interface.

One of the most notable additions to the command-line tooling is the introduction of Android Device Streaming directly within the CLI. While emulators are invaluable for initial development phases, testing applications on real physical hardware remains critical for catching subtle, hardware-specific or operating system-dependent issues. However, developers do not always have physical access to every specific device configuration or form factor they might need in their local offices. Android Device Streaming solves this challenge by granting remote access to an extensive fleet of real physical devices. Through the Android CLI, an AI agent can now leverage Android Device Streaming from virtually anywhere, bridging the gap between local development workflows and cloud-hosted hardware infrastructure.

This integration allows agents to interact with physical devices over a secure ADB over SSL connection, treating remote hardware as if it were physically plugged into the developer’s workstation via USB. This powerful capability enables development teams to spin up remote devices, deploy fresh builds, collect detailed logs and system traces, and even capture headless screenshots directly through the terminal environment. To begin utilizing this feature, developers simply need to link their project, instruct their agent to query the list of available remote devices, and select the specific target hardware required for their testing scenario. This addition significantly streamlines quality assurance and debugging workflows for teams relying heavily on command-line and agent-driven development paradigms.

Beyond infrastructure and device management, Google is addressing a fundamental limitation of modern artificial intelligence: the gap between an LLM’s default training data and the rapidly changing standards, APIs, and best practices of specific platform ecosystems. To solve this, the company continues to expand its official Android skills repository. Android skills consist of structured instructions contained within standardized markdown files designed to ground AI agents using verified guidance directly from the official developer portal. Instead of relying solely on a model’s static training cutoff date, these skills inject fresh data, precise API references, code samples, and modern architectural patterns directly into the agent’s context window.

With over twenty distinct skills currently available in the repository, developers can equip their AI agents to handle increasingly complex and specialized software engineering challenges. The development and evaluation of these skills follow rigorous internal methodologies, ensuring high reliability and effectiveness across diverse projects. Managing these resources across multiple codebases and individual agent directories is handled smoothly through the Android CLI. Because these skills are engineered to be environment-agnostic, they function seamlessly across a wide variety of setups. Whether a developer is writing code inside Android Studio, utilizing specialized terminal environments, or pairing third-party coding agents such as Claude and Codex, the Android skills provide consistent, reliable guidance across the entire development stack.

To illustrate the practical impact of these specialized instructions, Google highlighted the Wear Compose Material 3 skill in a detailed technical spotlight. Building software for Wear OS devices presents unique design and engineering challenges that differ significantly from traditional mobile development. Developers must account for round viewports, rotary input mechanisms, specialized ambient display modes, and aggressive power consumption constraints. Additionally, Wear OS development requires specific layout paradigms, such as prioritizing the TransformingLazyColumn component and properly utilizing AppScaffold and ScreenScaffolds containers. Without explicit, platform-specific guidance, standard large language models frequently lack the nuanced understanding required to implement these patterns correctly, often missing the details that make wearable applications polished and performant.

The Wear Compose Material 3 skill was created precisely to assist agents in overcoming these hurdles by providing deep contextual knowledge regarding modern Wear OS development standards. Early adopters across the industry have already reported significant productivity gains after integrating this skill into their daily workflows. For instance, the engineering team at FotMob adopted the skill to tackle several intensive modernization tasks on their existing Wear OS application. Their migration efforts included updating multiple legacy lists to use the modern TransformingLazyColumn component, correctly applying ScreenScaffold content padding, incorporating ListHeader titles, and implementing SurfaceTransformation effects on cards and buttons.

Furthermore, the team utilized the skill to standardize theme typography and generate accurate Wear-specific UI previews. The resulting codebase changes compiled successfully on the first attempt and were rigorously verified on emulators for smooth scrolling, rotary input behavior, edge morphing, and right-to-left language support. This successful migration allowed the FotMob engineering team to entirely eliminate their legacy wrapper code, as well as a significant amount of custom rotary and focus boilerplate that had previously required manual maintenance.

Importantly, the skill successfully caught several subtle architectural mistakes that the underlying AI model would have otherwise missed, such as forgetting to forward the critical contentPadding from the ScreenScaffold into the underlying list, and enforcing proper theme typography over hardcoded pixel measurements. Reflecting on the efficiency and accuracy of the process, Roy Solberg, an Android Tech Lead at FotMob, noted that a task involving eight migrated lists and the removal of extensive custom rotary code was successfully accomplished using a single skill over the course of a single afternoon. Through these comprehensive tooling updates and the continued expansion of grounded AI skills, Google aims to provide a robust, flexible foundation for the next generation of Android application development.