Android Studio is becoming a more open environment for AI-assisted development. Google has introduced a new “Bring Your Own Agent” (BYOA) feature that lets developers connect third-party coding agents directly to Android Studio, including Anthropic’s Claude Agent, OpenAI’s Codex, and Google’s Antigravity.

The feature is rolling out in preview with the latest Android Studio Canary release, giving developers a way to use the coding agent and AI plan of their choice without leaving their Android development environment.

Google says the move builds on Android Studio’s previous shift toward supporting a wider range of AI models. Rather than requiring developers to use a single built-in AI experience, the new system is designed around the idea that different developers and teams may already have preferred coding agents, enterprise AI setups, or custom workflows.

Android Studio becomes an agent hub

The new BYOA feature uses the Agent Client Protocol (ACP) to connect compatible coding agents with Android Studio. This allows an agent to receive information about the Android project, including the project graph, build configuration, and relevant platform details.

That context is important because an AI coding agent working on an Android project needs more than the contents of an individual source file. Understanding how the project is structured and built can help the agent identify which files matter for a particular task instead of consuming tokens on unrelated information.

Google says this approach is intended to improve token efficiency, reduce latency, and produce more relevant responses from the connected agent.

The integration also means developers can move between an AI agent and Android Studio’s existing development tools without having to completely switch workflows. Google describes this as seamless workflow continuity, allowing developers to keep their existing context while moving between agent-driven tasks and the IDE’s Android-specific capabilities.

Claude, Codex and Antigravity are supported

At launch, Google is highlighting three agents: Anthropic’s Claude Agent, OpenAI’s Codex, and Google’s Antigravity.

Developers can connect these agents from the agent window and sign in using the relevant account or provide an API key, depending on the service. Google also says additional ACP-compatible agents can be added through Android Studio’s agent registry under Settings > Tools > AI > Agents.

That makes the feature more significant than simply adding a few new AI assistants to Android Studio. The broader goal is to make the IDE compatible with an expanding ecosystem of coding agents.

For developers already paying for an AI service, that means they can potentially bring the subscription they already use into their Android workflow instead of maintaining a separate AI setup just for Android Studio.

Google also notes that developers can switch between agents. If one agent reaches its usage limit or isn’t delivering the desired results for a particular task, another connected agent can take over.

Agents can use Android Studio’s tools

The biggest advantage of the integration may be the tools available to an agent once it is working inside Android Studio.

Google says connected agents can read, write, and edit project files, execute shell commands, run tests, and search the web. Agents can also delegate work to subagents for tasks such as code review or testing.

More importantly for Android development, Google is providing native access to Android-specific tooling.

That includes build diagnostics, Jetpack Compose Previews, Android SDK tools, and Android Emulator control. An agent can therefore do more than generate a piece of Kotlin code and hand it back to the developer. It can work with the project, build it, test it, diagnose problems, and interact with the emulator as part of a larger workflow.

Google is also adding granular permissions so developers can decide how much autonomy an agent has. Routine operations can be allowed to run automatically, while potentially risky actions can require the developer’s approval.

The agents can also maintain longer-running sessions and automatically load project skills, slash commands, and memory, according to Google.

Gemini is still part of Android Studio

The new approach does not mean Google is abandoning its built-in Gemini experience.

Google says developers can continue using Gemini directly through Android Studio. However, for developers who want access to Google’s latest Gemini models and higher AI usage quotas, Google recommends using the Antigravity agent.

The company says Antigravity can provide access to models such as Gemini Flash 3.8, while developers can sign in using a Google AI Pro or Ultra plan or use a Gemini API key with pay-per-token pricing. Organizations using Gemini Enterprise can continue using either the built-in agent or Antigravity, with Google saying those deployments retain the security and privacy benefits of Google Cloud.

That gives Android developers several ways to use Google’s AI alongside the new third-party options rather than forcing everyone into one path.

What this means for Android developers

The interesting part of BYOA is less about any individual coding agent and more about where AI development environments are heading.

Coding agents are increasingly becoming capable of handling multi-step development tasks rather than simply answering programming questions. An agent can inspect a codebase, make changes across multiple files, run a build, investigate errors, execute tests, and iterate on the implementation.

Android Studio is now positioning itself as the environment where those agents can perform that work with access to Android-specific context and tools.

That could also make switching between AI providers considerably easier. A developer might use Codex for one task, Claude for another, and Google’s Antigravity for Gemini-powered workflows without having to abandon the Android Studio environment each time.

For teams, the flexibility could be particularly useful. Different developers may already have different AI subscriptions or organizational requirements, while enterprise teams may want to connect approved agents to an existing development environment.

Google is starting the feature as a preview rather than presenting it as a finished experience. BYOA support is rolling out with Android Studio Rabbit 2 in the Canary channel, with Claude Agent, Codex, and Antigravity among the initial supported agents. Developers can install the latest Canary build, open the agent window, select an available agent, and sign in or provide an API key.

The larger shift is pretty straightforward: Android Studio no longer needs to be the place where developers choose their AI agent. Increasingly, it can be the place where developers bring one.