Google Opens Android Studio to Third-Party AI Agents and Debuts Gemini Nano 4 APIs
At Android Dev Summit, Google introduced third-party AI agent support for Android Studio, new token-efficient CLI workflows, Gemini Nano 4 APIs, and the XREAL AURA smart glasses.
Google is restructuring its mobile development ecosystem around autonomous AI agents. At this year's Android Dev Summit, the company announced major updates across Android Studio, command-line utilities, system-level agent APIs, and augmented reality hardware. The centerpiece of the announcement is a shift away from closed toolsets toward an open environment that lets third-party AI models plug directly into the Android development lifecycle.
With the release of Android Studio Rabbit 2, Google is introducing a "Bring Your Own Agent" (BYOA) architecture. Built on the open Agent Client Protocol (ACP), the IDE removes previous restrictions that limited developers to Google's proprietary assistants. Through an integrated marketplace, engineers can now install and concurrently run over 40 coding agents, including Anthropic's Claude Agent, OpenAI's Codex, and Google Antigravity. These models are granted deep access to native IDE capabilities, including Logcat, the emulator, project structures, and symbol resolution.
Smarter developer tools and token-efficient workflows
To address the code quality issues that often accompany AI-generated pull requests, Google added Change Review. This verification utility analyzes pending code edits, explains changes step by step, and flags potential bugs. In the Android Studio Canary channel, developers also gain access to an automated migration assistant that translates legacy XML layouts into Jetpack Compose. Concurrently, Google is previewing Lightbuild, a declarative build system using a YAML schema intended for both human and AI maintenance.
For developers managing autonomous agents in terminals and CI environments, Google expanded the Android CLI tool into an official plugin for Claude Code, Codex, and Google Antigravity. A notable addition allows agents to render Jetpack Compose previews straight inside the terminal without booting an emulator. Internal tests from Google indicate that terminal-rendered previews allow agents to complete UI tasks 77 percent faster while cutting token consumption by up to two-thirds.
Google also integrated the open-source LeakCanary memory detection library into both the CLI and Android Studio Profiler. Heap dumps are processed on the host machine rather than the connected mobile device to prevent performance bottlenecks. In an A/B test running Anthropic's Claude Opus 5.5, the CLI integration helped the agent resolve memory leaks nearly twice as fast while lowering token usage by approximately 25 percent. Beyond debugging, agents can now leverage more than 30 curated "Android Skills" covering areas such as Play Store policy compliance, permissions security, and responsive layouts for foldables and Googlebooks.
Local agents, MCP, and Android XR hardware
Outside of the IDE, Google is rolling out tools to turn Android applications into agent-accessible resources. ML Kit is receiving an alpha Agentic API driven on-device by Gemini Nano 4, handling chat histories, safety filters, and execution loops locally. To unify how agents invoke app logic, Google adopted the Model Context Protocol (MCP) within its beta AppFunctions API. Building on this layer, the new Agent-to-UI (A2UI) preview for Jetpack Compose enables models to respond to users using dynamic, interactive app components rather than static text outputs.
Through the Connected Apps Platform preview, these app actions can also surface in the Gemini app and Google Search. Google confirmed that over 50 enterprise integrations, including Adobe, Peloton, and Experian, are active in the Gemini mobile app, with web search rollouts planned soon.
On the hardware front, Android XR reached stability alongside Jetpack Compose 1.13 and the Jetpack XR SDK. Google partnered with XREAL to launch the XREAL AURA, which XR chief Shahram Izadi identified as the first wired smart glasses running Android XR. Weighing under 95 grams, the glasses feature transparent, dimmable lenses with a 70-degree field of view, powered by a tethered computing puck with a Qualcomm Snapdragon Reality Elite processor. Integrated with Gemini, the glasses can analyze physical surroundings, arrange up to five virtual application windows, or mirror laptop screens. The hardware is priced at $1,099 in the US, Canada, and South Korea, with shipments beginning in mid-November.
What it means for developers
Google's updates signify an important strategic pivot: Android Studio is no longer a walled garden for Google's own AI products. By adopting open standards like ACP and MCP, developers gain the flexibility to pick the best model for specific software engineering tasks, whether that involves debugging memory leaks with Claude or refactoring layouts with Codex.
Token economics are also receiving overdue attention. As coding agents become regular participants in CI pipelines, features like terminal Compose previews and offloaded heap analysis directly reduce API bills. For developers looking to experiment with and benchmark different providers for their agent setups, platforms like Apixo (https://apixoai.online) offer cheap pay-per-token API access to top AI models like Claude, GPT, and Gemini through a single key.
Finally, the combination of Gemini Nano 4, MCP, and A2UI shifts the target for Android app design. Instead of simply building screens for human taps, mobile engineers will increasingly need to expose structured functions and modular UI components that both local and cloud-based agents can navigate on the user's behalf.
Source: Google opens Android development to third-party AI agents and unveils new Gemini APIs — The Decoder. Written by the Apixo team from that report.
One key for Claude, GPT, GLM, DeepSeek and more. Pay per token with crypto.
Get your API key
