Posted by Sandhya Mohan, Senior Product Supervisor and Trevor Johns, Developer Relations Engineer
We’re excited to announce that Android Studio Otter 3 Function Drop is now secure! This feature-packed launch brings an enormous replace to your agentic workflows in Android Studio, and gives you extra flexibility and management for the way you utilize AI that will help you construct Android apps.
- Convey Your Personal Mannequin: Now you can use any LLM to energy the AI performance in Android Studio.
- Agent Mode Enhancements: Now you can extra simply have Agent Mode work together along with your app on gadgets, evaluation and settle for instructed modifications, and have a number of conversations threads.
- Run person journey exams utilizing pure language: with Journeys in Android Studio.
- Allow Agent Mode to connect with extra instruments: together with the power to connect with distant servers through MCP.
- Construct, iterate and check your UI: with UI agentic experiences in Android Studio.
- Construct deep hyperlinks utilizing pure language: with the brand new app hyperlinks assistant.
- Debug R8 optimized code: with Automated Logcat retracing.
- Simplify Android library modules: with the Fused library plugin.
Right here’s a deep dive into what’s new:
Convey Your Personal Mannequin (BYOM)
Each developer has a singular workflow when utilizing AI, and totally different firms have totally different insurance policies on AI mannequin utilization. With this launch, Android Studio now brings you extra flexibility by permitting you to decide on the LLM that powers the AI performance in Android Studio, providing you with extra management over efficiency, privateness, and value.
Use a distant mannequin
Now you can combine distant fashions—equivalent to OpenAI’s GPT, Anthropic’s Claude, or an identical mannequin—straight into Android Studio. This lets you leverage your most popular mannequin supplier with out altering your IDE. To get began, configure a distant mannequin supplier in Settings by including your API endpoint and key. As soon as configured, you’ll be able to choose your customized mannequin straight from the picker within the AI chat window.
Enter the distant mannequin supplier data.
Use an area mannequin
In case you have restricted web connectivity, strict information privateness necessities, or a need to experiment with open-source analysis, Android Studio now helps native fashions through suppliers like LM Studio or Ollama. Whereas Gemini in Android Studio stays the default suggestion—tuned particularly for Android improvement with full context consciousness—if in case you have a selected mannequin choice, Android Studio helps it.
Mannequin picker in Android Studio.
An area mannequin gives an alternative choice to the LLM assist constructed into Android Studio, and usually requires important native system RAM and onerous drive house to run nicely. Nevertheless, Gemini in Android Studio supplies one of the best Android improvement expertise as a result of Gemini is tuned for Android and helps all options of Android Studio. With Gemini, you’ll be able to select from a wide range of fashions on your Android improvement duties, together with the no-cost default mannequin or fashions accessed with a paid Gemini API key.
Use your Gemini API key
Whereas Android Studio consists of entry to a default Gemini mannequin with beneficiant quotas without charge, some builders want extra. By including your Gemini API key, Android Studio can straight entry all the newest Gemini fashions obtainable from Google.
For instance, this lets you use the newest Gemini 3 Professional and Gemini 3 Flash fashions (amongst others) with expanded context home windows and quota. That is particularly helpful for builders who’re utilizing Agent Mode for prolonged coding classes, the place this extra processing energy can present larger constancy responses.
Agent Mode enhancements
Run your app and work together with it on gadgets
Agent Mode can now deploy an utility to the related machine, examine what’s at the moment proven on the display, take screenshots, verify Logcat for errors, and work together with the working utility. This lets the agent show you how to with modifications or fixes that contain re-running the appliance, checking for errors, and verifying {that a} explicit replace was made efficiently (for instance, by taking and reviewing screenshots).
Agent mode makes use of machine actions to deploy and confirm modifications.
Discover and evaluation modifications utilizing the modifications drawer
Now you can see and handle all modifications made by the AI agent utilizing the modifications drawer. When the agent makes modifications to your codebase, you’ll be able to see the recordsdata that have been edited in Information to evaluation. From there, you’ll be able to maintain or revert the modifications individually or all collectively. Click on a person file within the drawer to see the code diff within the editor and make refinements if wanted. With the modifications drawer, you’ll be able to maintain observe of edits made by the agent throughout your chat and revisit particular modifications with out scrolling again by your dialog historical past.
See all of the recordsdata that the agent has proposed edits to within the modifications drawer.
Settle for a change so as to add it to the modifications drawer.
Handle a number of dialog threads
Now you can manage your conversations with Gemini in Android Studio into a number of threads. This allows you to create a brand new chat or agent thread when it’s worthwhile to begin with a clear slate, and you’ll return to older conversations within the historical past tab. Utilizing separate threads for every distinct process can enhance response high quality by limiting the scope of the AI’s context to solely the subject at hand.
See prior conversations within the “Current Chats” tab.
Your dialog historical past is saved to your account, so if it’s important to signal out or swap accounts you’ll be able to resume proper the place you left off once you come again.
Journeys for Android Studio
Journeys for Android Studio.
These pure language directions are transformed into interactions that Gemini performs straight in your app. This not solely makes your exams simpler to write down and perceive, but additionally lets you outline advanced assertions that Gemini evaluates primarily based on what it “sees” on the machine display. As a result of Gemini causes about how to realize your targets, these exams are extra resilient to delicate modifications in your app’s format, considerably decreasing flaky exams when working towards totally different app variations or machine configurations.
Journeys for Android Studio.
Help for distant MCP servers
Connect with the Figma distant MCP server in Android Studio Settings.
Rapidly add a display to your app utilizing the Figma distant MCP server.
Supercharge your UI improvement with Agent Mode
Create new UI from a design mock
Speed up your preliminary UI implementation by producing Compose code straight from a design mock. Merely click on Generate Code From Screenshot in an empty Preview panel, and Gemini will use the picture to generate a beginning implementation, saving you from writing boilerplate from scratch.
Generate code from a screenshot in an empty Preview panel.
Instance turning design into Compose code.Match your UI with a goal picture
Upon getting an preliminary implementation, you’ll be able to iteratively refine it to be pixel-perfect. Proper-click your Compose Preview and choose AI Actions > Match UI to Goal Picture. Add a reference design, and the agent will counsel code modifications to make your UI match the design as carefully as doable.
Instance of utilizing “Match UI to Goal Picture”Iterate in your UI with pure language
For extra particular or artistic modifications, right-click in your preview and use the AI Actions > Change UI. This functionality now leverages Agent Mode to validate the outcomes, making it extra highly effective and correct. You should use pure language prompts like “change the button coloration to blue” or “add padding round this textual content,” and Gemini will apply the code modifications immediately.
Instance of utilizing “Change UI”
Discover and repair UI high quality points
Verifying your UI is high-quality and extra accessible is a vital remaining step. The AI Actions > Repair all UI verify instrument audits your UI for frequent issues, equivalent to accessibility points. The agent will then suggest and apply fixes to resolve the detected points.
Entry level to set off “Repair all UI verify points”
You too can discover the identical performance by utilizing the Repair with AI button in Compose UI verify mode:
“Repair with AI” in UI Examine mode
The options talked about above are additionally accessible by the toolbar icon within the Preview panel:
Second entry level to UI improvement AI options
Past iterating in your UI, Gemini additionally helps streamline your improvement surroundings.
To speed up your setup, you’ll be able to:- Generate Compose Previews: This characteristic is now enhanced by Agent Mode to supply extra correct outcomes. When working in a file that has Composable features however no @Preview annotations, you’ll be able to right-click on the Composable and choose Gemini > Generate [Composable name] Preview. The agent will now higher analyze your Composable to generate the mandatory boilerplate with right parameters, to assist confirm {that a} efficiently rendered preview is added.
Entry level to generate Compose Preview
- Repair Preview rendering errors: When a Compose Preview fails to render, Gemini can now analyze the error message and your code to search out the foundation trigger and apply a repair.
Utilizing “Repair with AI” on Preview render error
App Hyperlinks Assistant
The App Hyperlinks Assistant now integrates with Agent Mode to automate the creation of deep hyperlink logic, simplifying one of the time-consuming steps of implementation. As an alternative of manually writing code to parse incoming intents and navigate customers to the right display, now you can let Gemini generate the mandatory code and exams. Gemini presents a diff view of the instructed code modifications on your evaluation and approval, streamlining the method of dealing with deep hyperlinks and making certain customers are seamlessly directed to the proper content material in your app.
To get began, open the App Hyperlinks Assistant by the instruments menu, then select Create Applink. Within the second step, Add logic to deal with the intent, choose Generate code with AI help. If a pattern URL is offered, enter it, after which click on Insert Code.
App Hyperlinks Assistant
Automated Logcat Retracing
Debugging R8-optimized code simply grew to become seamless. Beforehand, when R8 was enabled (minifyEnabled = true in your construct.gradle.kts file), it might obfuscate stack traces, altering class names, strategies, and line numbers. To seek out the supply of a crash, builders needed to manually use the R8 retrace command line instrument.
Beginning with Android Studio Otter 3 Function Drop with AGP variations 8.12 and above, this additional step is not needed. Logcat now routinely detects and retraces R8-processed stack traces, so you’ll be able to see the unique, human-readable stack hint straight within the IDE. This supplies a much-improved debugging expertise with no additional work required.
Logcat now routinely detects and retraces R8-processed stack traces
Fused Library Plugin: Publish a number of Android libraries as one
Get began
Able to dive in and speed up your improvement? Obtain Android Studio Otter 3 Function Drop and begin exploring these highly effective new options at the moment!
As at all times, your suggestions is essential to us. Examine recognized points, report bugs, and be a part of our vibrant neighborhood on LinkedIn, Medium, YouTube, or X. Let’s construct the way forward for Android apps collectively!