Google Unveils Android CLI for Third-Party AI Agents

Google releases Android CLI 1.0 for third-party AI agents.

Google has released the stable version of Android CLI 1.0, opening the tool for third-party AI agents. The company announced the launch at the I/O 2026 conference.

The update components also include:

• Android Skills: a GitHub repository with Markdown instructions. These are modular rule sets that help neural networks (Gemini, Claude Code, and others) better understand the specifics of Android development, from navigation to migration to Compose;
• knowledge base: a specialized database accessible via the android docs command. It provides AI agents with up-to-date context from Android, Firebase, and Kotlin documentation, even if the model’s training data is outdated.

Previously, Android CLI was primarily used with Gemini CLI and Google Antigravity. Now, the tool supports external agents like Claude Code or Codex. According to Google, this approach allows for app creation three times faster.

Using Android CLI with Android skills reduces LLM token consumption by approximately 70% compared to regular chatbots lacking platform context.

Integration with Android Studio

In version 1.0, the tool has gained a close connection with the IDE. Through a new set of android studio commands, AI agents have access to:

• error analysis and symbol declaration searches;
• rendering Compose Preview;
• checking current dependency versions.

Effectively, Android CLI has become a software layer over Android Studio, providing models with “semantic” access to project structure and platform resources.

Google emphasized that development is not limited to its own models. The documentation explicitly mentions support for Claude and GPT. Android Studio itself now allows connecting any models, including the local Gemma 4.

Android CLI has been integrated into the Antigravity 2.0 platform and Google AI Studio. The latter has gained native development capabilities, including:

• automatic application code generation;
• built-in emulator;
• deployment on physical devices and publishing to Google Play (internal tests).

To automate complex tasks, a Journeys system has been implemented, allowing for the description of action sequences (e.g., “go to the cart and click pay”) in simple terms.

Despite the expansion of terminal capabilities, Google maintains Android Studio as the main hub for complex debugging and profiling, positioning the CLI as a bridge between AI and the professional development environment.

A preliminary version of Android CLI is already available for download. Developers can also use the android sdk install command for quick setup of necessary environment components.

Back in May, the corporation tightened its policy on the use of search AI tools following a public experiment by BBC journalist Thomas Germain.