Google opens Android Studio to rivals, and that's the smart move

What do you do when the developers who build your platform have already decided which AI they like, and it isn't yours? If you're Google, you stop arguing and hand them the keys. That, in our view, is the real story behind Android Studio's new openness.

What changed, briefly

At its Dev Summit, Google presented Android Studio Rabbit 2, and heise reports that, alongside new agentic features, it lets developers use non-Google AI. Digital Trends says the IDE now supports over 40 AI coding agents, Claude and Codex among them.

Google's own developer blog frames it as a second step: last year Android Studio opened up to any AI model, and now it supports your choice of coding agents. So this is a trajectory, not a one-off U-turn.

Google is conceding a battle to win the platform

Look at the sentence Google chose. Its blog says it's important for developers to build Android apps "in the way that works best for you and your team." That is the language of a company that has read the room. Many developers already use ChatGPT, Claude and GitHub Copilot, as Google's own documentation admits. Forcing them to switch tools to work on Android would have been a tax on the platform, not on the competitors.

We think that's the key insight. Android's strength is the number of apps and developers on it. An IDE that annoys those developers to promote Gemini would be trading something valuable for a vanity metric. Letting Claude and GPT in costs Google a bit of pride and probably protects the thing that matters: that building for Android stays the path of least resistance.

It's also a quiet reversal of the usual playbook. Platform owners typically use their tools to bind developers to their own services. Here Google is betting that being the neutral workbench is worth more than being the exclusive one.

The catch: defaults decide more than menus

Choice on paper isn't choice in practice. Most people never leave the default, and a developer who opens Android Studio and finds Gemini pre-wired, pre-suggested and first in every menu will probably use it. The documentation we have doesn't tell us how Gemini is positioned in the interface, so we can't say whether Google is playing fair. That is precisely the thing to watch.

There's also a real trade-off that Google is honest about. Its pages for both remote and local models warn that by using a third-party model, you agree to send your code and other input data to the provider, so you should review the provider's terms. For a freelancer, fine. For a company with proprietary code, that line needs a conversation with legal. Google also supports local models running on your own machine, which we think will appeal to teams that care about privacy more than raw capability.

And a word on security, since agents are only as trustworthy as the tooling around them. Golem reported this week that attackers injected malicious code into the official VS Code extension for Claude Code. That's a different product and a different IDE, but it's a reminder that opening your editor to many agents widens the surface you have to vet. Openness is good; blind trust isn't.

Our verdict

This is a smart, pragmatic concession, and developers are the clear winners: they keep their workflow and get to compare models on their own code. Anthropic and OpenAI gain distribution inside the IDE that every Android developer opens. Google gives up exclusivity and keeps the relationship.

The only way it goes wrong is if openness becomes a fig leaf, with Gemini holding every privileged default while the rivals sit in a sub-menu. If you build Android apps, our advice is simple: try at least two agents on a real task, check where your code goes before you commit to one, and don't let the default choose for you. Better apps will come from developers who pick deliberately, and Google has finally let them.