# Android Studio: Pick Your Own Coding Agent

At Android Dev Summit, Google said Android Studio now runs more than 40 coding agents of your choice. This follows its announcement through code review, terminal tools and performance analysis.

- Language: en
- Published: 2026-10-29
- Tags: app, android, android-studio
- URL: https://devs7ud.io/en/blog/android-dev-summit-pick-your-agent-en
- Translation (ko): https://devs7ud.io/blog/android-dev-summit-pick-your-agent.md

## Pick your own coding agent

At Android Dev Summit, Google said Android Studio now lets you choose from more than 40 coding agents. Any agent that follows ACP, the Agent Client Protocol, can be installed and used inside the IDE.

This post follows the wrap-up on the Android Developers Blog. I have not used these tools myself, so please read it as a note on a fresh announcement, not as a confirmed trend.

## Bring your own agent

In Settings under Tools, AI and Agents you browse the marketplace and install agents. You can keep several and switch between them as you work. New agent features arrive the moment they ship, and the agent gets everything Android Studio knows about your project.

Google Antigravity comes in the same way. Log in and you get new Gemini models as they launch, Google says. Honestly, which company's agent you use is no longer a reason to pick an IDE.

![Claude as an agent inside Android Studio](https://developer.android.com/static/blog/assets/BYOA_Claude_no_rounding_64d9b82578_Z2b5Xun.webp "An agent like Claude, inside the IDE")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

![Android Studio's Change Review](https://developer.android.com/static/blog/assets/change_review_177f39b007_2qhMik.webp "Reviewing what an agent changed")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

![Android Dev Summit Tools and Performance spotlight image](https://developer.android.com/static/blog/assets/ABL_146_ADS_Spotlight_Tools_and_Performance_47f52bd6b1_Z1qi2aR.webp "The summit announcement")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

## Reading what agents changed

When an agent makes a big change, reading it becomes the job. The new Change Review in Android Studio starts from pending agent changes or the Git changes window and produces a report of potential issues.

Android Bench 2.0 helps you judge which model fits Android work. It is an LLM leaderboard that scores models and agent tools on long, realistic Android tasks.

## Emulators and moving to Compose

Emulators for XR glasses, audio glasses and display glasses are stable and can be created in Android Studio. You can dress the environment with your own 360-degree photos or a connected camera feed to test real places from your desk.

Moving XML Views to Compose has always been manual work, and the Migration Assistant takes most of it. Right-click an XML layout and choose AI, Migrate to Compose, or run it across the project. A progress timeline shows each step. It is in the latest Android Studio Canary.

![An emulator with a custom 360-degree environment](https://developer.android.com/static/blog/assets/emulator_environment_86a016acf3_ZrhT3l.webp "An emulator dressed with my own photos")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

![The Migration Assistant moving XML Views to Compose](https://developer.android.com/static/blog/assets/views_to_compose_0ae4c582e3_BqpJm.webp "From XML to Compose")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

![android device hub showing devices](https://developer.android.com/static/blog/assets/adh_screenshot_b80681b058_1WOQTR.webp "android device hub")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

## Android in the terminal too

Android CLI lets any agent, in any terminal or CI, do Android work. It now streams real devices: you borrow a physical device from a Google device lab and install and test as if it were plugged in. Device Hub shows all your devices at a glance.

The android preview command renders Compose Previews without an emulator, so an agent can check its own UI changes. In Google's internal tests, agents using it finished UI tasks 77% faster with up to 3x fewer tokens. There are official plugins for Claude and Codex too.

| Command | What it does |
| --- | --- |
| android device remote | Connect to a real device in a Google device lab |
| android device hub | See local and remote devices at a glance |
| android preview | Render Compose Previews without an emulator |
| android profiler leakcanary | Analyze memory leaks |
| android signals | Look into crashes and ANRs in production |
| android skills list | List the official skills |

_Source: Android Developers Blog_

## Finding memory leaks together

LeakCanary is now part of Android CLI. android profiler leakcanary analyzes heap dumps on your computer instead of the device. Google says the most capable models finished the job almost 2x faster using up to 25% fewer tokens. In the keynote, Claude Code got a list of critical user journeys and one prompt, then found and fixed leaks across several devices in parallel and re-ran the journeys to confirm.

In Android Studio, Fix with AI inside the Profiler reviews a leak and suggests a fix. It is in the Canary build, and the CLI command is coming soon.

![LeakCanary with and without Android CLI](https://developer.android.com/static/blog/assets/Leak_Canary_with_Android_CLI_and_without_db9752527d_11k207.webp "With and without Android CLI")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

![The keynote demo of Claude Code finding and fixing memory leaks](https://developer.android.com/static/blog/assets/cc_demo_2x_half_frames_720p_1277663323_Z1FOu5N.webp "Claude Code finding leaks in the keynote")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

![Fix with AI on a memory leak](https://developer.android.com/static/blog/assets/Leak_Canary_AI_Fix_Zoomed_in_b219dedb82_2tSPWQ.webp "Fixing a leak with Fix with AI")
_Image: [Android Developers Blog](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)_

## Lightbuild and official skills

Lightbuild is an early preview that makes build configuration data, not code: a strict YAML schema that people and agents can read and edit with confidence. Google calls it early and suggests new or experimental projects, while production apps stay on Gradle and AGP.

Models are not always trained on the latest Android APIs, so Google publishes official skills. There are now over 30 across more than 20 areas, each backed by evaluations. A play-policy-insights skill checks your app against Google Play policies before you submit.

## View

The striking part is the direction, not one feature. The IDE no longer picks the agent; it welcomes yours. Numbers like 77% or 2x are Google's internal results, and I do not know whether my own project would see the same speed. I would start with Change Review.

## Sources used here

- [What's new in Android developer tools and performance at Android Dev Summit (Android Developers Blog)](https://developer.android.com/blog/posts/android-dev-summit-26-developer-tools-performance)