Google brings device streaming and new AI skills to Android CLI

Google is expanding Android CLI with device streaming and new Android Skills, giving AI coding agents more ways to build, test and troubleshoot Android apps from the command line.

The update is aimed at extending AI-assisted development beyond code generation. Developers can use Android CLI to connect coding agents with Android-specific tools and workflows, while the latest additions give agents access to remote devices and more specialized development knowledge.

One of the main additions is device streaming, which allows agents to access remote Android devices during development. This means an agent can deploy an app, interact with it and inspect its behavior without requiring the developer to have the target hardware physically connected.

The capability could be particularly useful when testing apps across different Android form factors and device configurations. Instead of working only from source code, an agent can incorporate what happens when the app is actually running into its development workflow.

Google is also expanding Android Skills, which provide coding agents with instructions and context for specific Android development tasks.

The skills are designed to give agents access to current Android practices in areas where APIs and development patterns can change quickly. Developers can add relevant skills to their agent workflows rather than supplying the same technical context manually for each task.

Android CLI is designed to work with multiple coding agents rather than being tied exclusively to Google’s own AI models. This gives developers a common command-line layer for connecting agents with Android development capabilities.

The tool can also work alongside Android Studio, allowing agents to access additional project context and development functionality while developers continue using the IDE.

Another part of the workflow is Journeys, which lets developers describe app interactions in natural language. Agents can use these instructions to navigate and test an application, helping developers check whether specific user flows continue to work after code changes.

Combined with device streaming, this creates a development loop in which an AI agent can modify an app and then interact with the resulting build rather than stopping after generating code.