Google releases Agent Mode for Gemini in Android Studio

Google has released Agent Mode for Gemini in Android Studio, a feature now available in the Narwhal Feature Drop Canary release. Announced at I/O 2025, Agent Mode represents a functional shift from passive code suggestions to active task execution. It allows developers to describe high-level goals, which the system then translates into actionable plans that modify code across multiple files.

Unlike traditional AI assistants, Agent Mode allows developers to input prompts like “Fix build errors” or “Extract hardcoded strings to strings.xml”. The system creates a step-by-step plan using internal Android Studio tools to read, edit, and build the codebase. Developers can monitor progress, accept or reject individual changes, and provide feedback for iteration.

There is also an Auto-approve setting, where changes are applied without manual review, aimed at speeding up less critical workflows.

Agent Mode can be used with the default Gemini model, which includes a limited no-cost quota. However, developers can add a Gemini API key to enable Gemini 2.5 Pro, which supports a significantly larger context window—up to 1 million tokens. This allows for processing larger codebases and more complex tasks.

Developers on the business-tier subscription already have access to Gemini 2.5 Pro and do not need to configure an API key separately.

The release also introduces initial support for the Model Context Protocol (MCP). This allows Agent Mode to interface with external tools through configuration using a mcp.json file. One example includes linking with the GitHub MCP server to enable pull request generation directly from Android Studio. Additional integrations, such as memory and reasoning servers, are also supported.

Future versions are expected to expand MCP capabilities with additional protocols and resource access.

For mobile developers, Agent Mode can automate routine tasks like refactoring, screen implementation, and codebase cleanup. This could shift how teams handle time-consuming updates, freeing up resources for more strategic areas like ad placement testing, SDK maintenance, or layout adjustments.

Agent Mode may also affect development timelines, allowing faster iterations on UI and functional updates required by ad tech platforms or partner integrations.

Agent Mode is currently available in the Canary release of Android Studio Narwhal. Developers can activate it through the Gemini panel in the IDE. Use with a personal API key or access the full functionality via a business-tier license. Configuration for MCP support is available through the official documentation.