Google Opens Android Development Tools to Claude Code and Codex

Oct 28, 2026 693 views
Green circuit board

Android’s latest developer tools give coding agents access to more of the workbench: project context, device testing, performance data and production crash reports. That could let agents handle a longer stretch of the development cycle, from investigating a defect to checking whether a fix works.

Google is extending these capabilities through Android Studio and its command-line tools, while allowing developers to choose third-party agents. Matthew McCullough, vice president of product for Android, has highlighted that flexibility. The updates were announced Wednesday at Android Dev Summit. In a Wednesday post, Adarsh Fernando, senior product manager for Android developer tools, said the aim is to give developers and agents “the tools and knowledge to build high-quality Android apps faster.”

Agents get project context inside Android Studio

Android Studio’s Bring Your Own Agent feature now supports more than 40 coding agents that use the open Agent Client Protocol. The list includes Anthropic’s Claude Agent, OpenAI’s Codex and Google’s Antigravity. Google says the feature is stable in Android Studio Rabbit 2.

Developers can install multiple agents and switch among them without leaving the IDE. Connected agents can use Android Studio’s understanding of a project, including its structure, build setup and symbols, as well as tools such as Compose Preview, the emulator and Logcat. That context can spare an agent from piecing together the project through searches and terminal commands.

A new Change Review feature checks an agent’s pending edits for common problems and helps developers review larger changes before applying them. Google is also bringing Android tools to the terminal with official plugins for Claude Code and Codex. They support tasks such as creating projects, managing emulators, inspecting interfaces and accessing Android documentation.

Compose previews and device testing from the terminal

The new android preview command lets agents render Jetpack Compose previews in the terminal, without opening Android Studio or an emulator. Google reports that agents using the command completed UI tasks 77% faster and used up to three times fewer tokens. The company provided little detail about how it measured those results.

The CLI also adds android device remote, which gives agents access to physical devices in Google’s lab for testing. Taken together, these commands move more of the build and test loop into agent workflows, though the reported speed gains need more measurement detail to assess.

agents using it completed UI tasks 77% faster while using up to three times fewer tokens, though it shared little about how it measured those results.

Agents can trace leaks and production failures

Google is also connecting agents to performance investigations, memory leaks and crashes from deployed apps. In a demo, Claude Code used LeakCanary, an open source leak-detection library, to examine critical user journeys across several devices. It analyzed heap dumps through the CLI integration, traced leaks to code, applied fixes and reran tests.

Google says capable models completed these investigations nearly twice as fast while using up to 25% fewer tokens. A companion android signals command brings Android vitals and Firebase Crashlytics data into the terminal, giving agents access to production crash reports, application-not-responding events and memory issues.

The workflow can speed up diagnosis, but finding a possible bug doesn’t guarantee a sound fix. Anthropic’s scanner, for example, flagged some 29,000 possible vulnerabilities in open-source code while only a few hundred were ever fixed. And even code that appears to work can hide problems: reviewers found a race condition in a Linux kernel fix landed by Claude Code. Human review remains part of the process.

A companion android signals command now pipes Android vitals and Firebase Crashlytics data into the same terminal workflow

Android Bench expands agent evaluation

Android Bench 2.0 is Google’s new way to assess models and agent harnesses on more demanding Android tasks. Assignments include building an app from mockups and updating an existing interface. The benchmark checks whether the resulting apps run, and tracks completion time and cost. Some assignments are based on work that could take an Android engineer several days.

The practical shift is broader than a new IDE feature: agents can now draw on more of Android’s development and debugging environment, including signals from real devices and deployed apps. Android Bench may help compare how well they handle that work, while developers will still need to judge whether their changes are correct.

The post Google just handed its rivals the keys to Android appeared first on The New Stack.

Source: Amanda Caswell · thenewstack.io

Comments

Sign in to comment.
No comments yet. Be the first to comment.

Related Articles

Google just handed its rivals the keys to Android