Agent mode

Gemini in Android Studio's Agent mode is designed to handle complex, multi-stage development tasks that go beyond what you can experience by just chatting with Gemini. You can describe a high-level goal, and the agent will create and execute a plan, invoking the necessary tools, making changes across multiple files, and iteratively fixing bugs. This agent-assisted workflow empowers you to tackle intricate challenges, accelerating your development process.

Get started

To get started in Agent mode in Android Studio, follow these steps:

  1. Click Gemini in the side bar. Sign in and onboard if you need to.
  2. Select the Agent tab.
  3. Describe the task you want the agent to perform.

  4. As the agent goes through the steps to accomplish the task, you'll have the option to review and approve any changes.

    Agent mode waits for you to accept or reject a change.

  5. Optional: To automatically approve changes, select Agent options > Auto-approve changes.

    Auto-approve changes suggested by the agent.

Use cases

Here are some examples of use cases that the agent can help you with:

  • Fix build errors. When you ask the agent to fix a build error using a prompt like "Fix build errors in my project," it will apply a recommended fix, build the project to verify the solution, and iterate until the issue is resolved.

    The agent builds your project and assesses if there are errors.

  • Agent mode can add or update UI elements all on its own. For example, ask the agent to "Make dark mode the default in user preferences" and it'll search for the relevant files and suggest changes to accomplish the task. You can immediately preview the UI updates in the Running devices tool window.

  • It's helpful to have mock data when prototyping and testing your app. Instead of manually creating fake data yourself, you can give that task to the agent. Here we've asked the agent to "Add two more sessions to the fake data." It finds the relevant file and adds two more events to the mockSessions list.

Some other prompts to try:

  • "Write unit tests for <class> in <module>."
  • "Generate documentation for the currently open file."
  • "Change the app title from <current name> to <new name>."
  • "Resolve the null pointer exception."
  • "Refactor my code by moving the <composable name> composable to a new file. Ensure all imports are updated."
  • "Add a new button to the app's home screen called 'Follow' that takes you to a list of topics."
  • "In the <composable name> composable, reduce the padding of the <modifier name> modifier."
  • "Create a share button to share to social media."

How the agent works

In Agent mode, your prompt is sent to the Gemini API with a list of tools that are available. You can think of tools as skills: they include the ability to search for files, read files, search the text within files, use any MCP servers you have configured, and more.

When you give the agent a task, it comes up with a plan and figures out which tools are needed. Some of these tools might need you to grant permission before the agent can use them. Once you grant permission, the agent uses the tool to perform the necessary action and sends the result back to the Gemini API. Gemini processes the result of the action and generates another response. This cycle of action and evaluation continues until the task is complete.