
This year at Google I/O we are going beyond iterative changes, towards a fundamental shift in how apps are built. Our newest tools are built for the agentic era with features that boost productivity for you as an Android developer AND supercharge the AI agents you deploy in your codebase. So, whether you are building exclusively with AI or you prefer being the architect of every line of code, our tools will keep you ahead of the curve.
As we move from “AI-assisted” to “Agentic” development, we’re making it easier than ever to turn a spark of an idea into a high-quality production app with significantly less developer effort.
So what’s new with Android developer tools? We will cover 3 main areas in this blog:
-
Let your agent handle it: Whatever development task you are working on, the Android Studio agent can help: from planning the app architecture and design, to writing code, to unit testing and bug fixing.
-
Any AI provider, anywhere you build: In Android Studio, you can use any model and we even help guide you to the best performing ones. Choose any of the top remote models from Google, Anthropic, OpenAI, or if you need to run locally – Gemma 4 is our most capable and efficient local model! And with Android CLI, you can build Android apps faster and easier using the agents and developer environments of your choice.
-
As always, performance and quality remain top priorities: We continue to invest in the Android developer tools you love: from the Emulator, to Profilers, performance analyzers, and more!
1: Let your agent handle it
Agent skills
Android Studio now supports Agent Skills, modular instruction sets that ground LLMs in specialized workflows and domain-specific knowledge. By adding skills to your project, you can teach the agent to follow specific best practices, architecture patterns, or library workflows. This enables more accurate, context-aware code generation and automated skill activation for an appropriate task, ensuring the agent acts as an expert. We’ve bundled many of the top Android and Firebase agent skills in the latest Android Studio Canary build, so you can skip straight to building!
Skills in Agent ModeYou can create your own skill, or use Android CLI to install our official skills – a repository that covers some of the most common workflows that some Android developers and LLMs may struggle with. They help models better understand and execute specific patterns that follow our best practices and guidance on Android development, such as XML to Compose migration, Edge-to-edge, Navigation 3, and more. You can even build for Android XR, starting with a beautiful Display Glasses app with Jetpack Compose Glimmer. Official Android skills are automatically bundled with the latest Android Studio so the Agent is ready to build!
Build full-stack apps with Firebase in Agent Mode
Firebase services like Auth and Firestore databases can now be enabled directly within Agent Mode in Android Studio using the Agent Skills for Firebase. Your agent will be able to complete Firebase integration and configure backend services. This integration empowers you to build robust, full-stack Android applications without ever leaving your IDE!

Building a full-stack app with Firebase via Agent Mode
Parallel conversations
You can now run multiple conversations with Agent Mode in parallel. In one conversation, run tests and while you are waiting, you can kick off planning mode for a new feature in your app while using a third conversation thread to write documentation for your app. These improvements will save you time and improve your productivity.

Parallel conversations in Agent Mode
A more capable New Project Agent
Android Studio’s New Project Agent has evolved into a powerful full-stack development tool, utilizing a multi-step execution plan and an autonomous “generation loop” that self-corrects build errors and configures dependencies across multiple files. This advanced capability is significantly amplified by its new integration with Firebase Agent Skills, allowing developers to seamlessly build, debug, and deploy complete full-stack applications directly from a single prompt to final production.

Building an app with New Project Agent
Additionally, it now offers support for large screens. You can scaffold your project with layouts, navigation, and components optimized for tablets, foldables, and laptop devices from the get-go. It has additional logic to test your app on large-screen emulators if you have one enabled. Simply configure the required device in the Android Emulator and the Agent can test it out!

Build large screen apps for foldables and tablet devices
2: Any AI provider, anywhere you build
Build Android apps in Google AI Studio
Google AI Studio now features full Android app development capabilities. Users can generate new applications, preview them instantly via an embedded Android Emulator, and deploy them directly to physical devices using ADB over USB. Additionally, developers can publish straight to Google Play; AI Studio handles the app record creation, bundles the package, and uploads it to an internal testing track. For advanced development and production readiness, projects can be exported as a ZIP file and opened seamlessly in Android Studio. To get started, visit Google AI Studio today and start building!
Google AI Studio build mode with Android frameworkAndroid CLI helps you build faster, more efficiently with any agent
Android CLI enables you to build apps using any agent, LLM, and tool of your choice. Android CLI is designed to help AI agents build faster, and use less tokens when compared to only using generic LLM tools. By grounding agents with Android Knowledge Base and Android skills, you can now have your agent of choice follow the latest best practices across any coding environment.
Additionally, when using the latest Canary version of Android Studio Quail, Android CLI enables your agent to leverage powerful capabilities of the IDE, such as analyzing files for issues or finding symbol declarations. Google Antigravity 2.0 now offers official support for Android development with Android CLI.
And now in the latest Canary build, you can download and run Gemma 4 directly from the IDE, without needing to set up an external server.
Android Studio allows developers to bring any model they choose into the IDE for agentic AI assistance. Power your workflow with models like Gemini, GPT, and Claude or use a local model like Gemma 4. This flexibility offers developers greater control over performance, privacy, and cost.

