Google launches a major attack on Cursor AI and Copilot with Firebase Studio

At its Cloud Next conference, Google presented Firebase Studio, a comprehensive, cloud-based development environment with AI at its core. It extends the Firebase tool platform with AI workspaces in which developers can build complete full-stack applications, mobile apps and websites with AI chat. There are also AI-controlled tools for testing, data connection, documentation, and deployment.

In addition to a classic code editor, Firebase Studio offers a prototype agent that users can control entirely via natural language or images such as diagrams and UI sketches: The developers explain what kind of app they wanted and what it should look like. The AI then builds an application independently according to these specifications and shows a live preview. This can then be corrected or refined via chat. No work on the code is necessary here: No code.

In the classic code editor, developers will find over sixty templates on which they can base their projects, including for Go, Java, .NET, Node.js and Python Flask as well as frameworks such as Next.js, React, Angular, Vue.js, Android and Flutter.

AI access on all platforms

With Firebase Studio, Google is launching a major attack on competitors such as Cursor or GitHub Copilot (which now also processes graphics). The AI environment links various services, platforms, and interfaces and generates the UI, backend, frontend, mobile apps, API schemas and data connections in addition to the program logic. There are also additional functions for debugging, testing, documentation, deployment, and monitoring. Even after the release, developers can continue to work on the project in the Studio, expanding and modifying it.

Users have access to the tools of the previous Firebase platform for testing, messaging, authentication, data connection and app hosting, among other things. The latter can be called up directly via a button in the development environment for quick publication of your project, but is subject to a charge. The extensions of the Open VSX Registry can also be used.

A special agent for Android testing simulates user interactions and site visitors. It works through specific tasks (e.g., “Find a trip to Greece”) and provides error reports with user paths.

Existing code can be integrated at the beginning via repositories and deployment can be initiated at the end via Google Cloud Run or another infrastructure.

By and for agents

Additional coding agents can be found via the newly launched Code Assist Agents, for example for migration, adversarial tests or documentation. This offer requires a special registration, currently with a waiting list. The assistants will later also be available for VS Code.

For AI projects, Firebase Studio offers the functions of Genkit, which can be found as a template in the editor. It handles RAG, data connection, Model Context Protocol (MCP), multimodel control and communication between agents.

Firebase Studio works in a complete virtual machine that can be configured with the package manager Nix. In the preview phase, three workspaces (Editor or Prototype) are free, and thirty workspaces are free for participants in the Google Developer Program.

Payment is planned: pay as you go. Projects will then be subject to a charge if the user wants to use them. Further information on Firebase Studio can be found on the Google Cloud blog and in the documentation.

(who)