Google allows 40 third-party AI agents to operate on Android Studio

Google activates the use of third-party AI agents in Android Studio and delivers a series of APIs to run agents on devices and within the Android ecosystem. The company highlights measured productivity gains on the terminal side and announces operational partner integrations, while the XREAL AURA glasses, compatible with Gemini, are now available for order.

Timed Gains and Active Integrations

According to internal tests, displaying Compose previews in the terminal allowed agents to complete interface tasks 77% faster and use up to three times fewer tokens. A separate A/B test, conducted with Anthropic's Claude Opus 5.5 model, shows that the command-line tool enabled memory leak fixes almost twice as fast while reducing token usage by about 25%. Google also notes that there are already over 50 active partner integrations in the Gemini app, including Adobe, Peloton, and Experian, with a web search launch expected in the coming weeks. On the hardware side, the XREAL AURA glasses are priced at $1,099 and available for order in the United States, Canada, and South Korea, with shipments scheduled to begin in mid-November.

The Android Terminal Gains Dedicated Agent Commands

The Android CLI tool is now offered as an official plugin for Claude Code, OpenAI's Codex, and Google Antigravity. New commands allow for displaying Compose previews directly in the terminal, eliminating the need to open an emulator to visually control interface changes. The CLI also enables streaming of physical test device feeds from Google's labs to analyze device-specific bugs. The LeakCanary library has been added to the CLI and the Android Studio Profiler, with memory dump processing done on the computer rather than on the device to maintain testing speed. Debugging and performance analysis thus inherit additional features.

Google Launches Interfaces to Integrate AI Agents into Android

Google is offering interfaces designed to integrate agents into applications or connect apps to system agents. An alpha Agentic API, powered by Gemini Nano 4 and integrated with ML Kit, executes functions automatically, maintains a history of exchanges, orchestrates loops, and applies security filters. The beta AppFunctions API serves as a unified tool-calling layer via the Model Context Protocol, allowing applications to share functions with local or system agents. Meanwhile, the alpha Agent-to-UI interface for Jetpack Compose allows for responses in the form of interactive interfaces based on application components. To connect these services to Google Search and the Gemini app, an overview of the Connected Apps Platform enables applications to be attached via Android MCP or cloud MCP servers, managed from a single console.

Android Studio Adopts BYOA and AI Benchmark 2.0

With Android Studio Rabbit 2, the Bring Your Own Agent logic allows for the direct execution of over 40 coding agents in the IDE via the Agent Client Protocol. Developers can utilize Claude, Codex, or Google Antigravity, with access to project structure, symbol resolution, the emulator, and Logcat. Google is also releasing Android Bench 2.0, a ranking that measures model performance on complex development tasks, evaluating multi-day projects in terms of functionality, time spent, and cost. These developments are part of the system tool and API updates presented this year, lifting the exclusive dependency on in-house AI tools.

Stable XR Version and XREAL AURA Glasses Integrated with Gemini

The stable version of the Extended Reality development environment is now available, integrating the Jetpack XR SDK, XR emulators in Android Studio, and an Engine Hub compatible with Unity, Godot, and Unreal. The XREAL AURA glasses, introduced simultaneously, can display up to five virtual windows, weigh less than 95 grams, and feature transparent lenses with adjustable brightness, offering a 70-degree field of view. Processing and power supply are handled by an external module, which also serves as a touchpad and uses Qualcomm's Snapdragon Reality Elite chip. Users can display up to five Android applications simultaneously or mirror a laptop screen via USB-C. Gemini is integrated, with real-world environment perception and virtual screens to organize windows, guide a gaming session, or highlight physical objects. The glasses are priced at $1,099.

A Dev Summit Focused on AI Agents, from Studio to APIs

The Dev Summit emphasizes agents, with integrations for models from Anthropic and OpenAI, new APIs for agents operating on smartphones, and a hardware announcement regarding glasses. Updates to Jetpack Compose 1.13 and the AppState model aim to adapt to large screens. On the local execution side, an Agentic API powered by Gemini Nano 4 and the Model Context Protocol serves as the backbone for new agent pathways. Android Studio is opening up to third-party agents, such as Claude and Codex, while terminal tools are being expanded to reduce development time and token usage. To ensure reliable generation, Change Review inspects modifications, creates bug reports, and clarifies each step, an assistant automatically migrates XML layouts to Jetpack Compose, and Google is testing Lightbuild, a declarative system based on YAML accessible to both humans and agents.