Antigravity Agent Is Now Built Into Google AI Studio — Your Guide to the New Firebase Workflow

Firebase Studio Is Gone — But Where Did Everything Go?

If you recently got the Firebase Studio sunset notice and aren't sure what to do with your projects, here's the short answer: everything moved into Google AI Studio 2.0.

In March 2026, Google announced that Firebase Studio's functionality would be absorbed into AI Studio's new Build mode — powered by the Antigravity coding agent, the same engine behind the standalone Antigravity desktop IDE. The result is a browser-based development environment where you can build full-stack, Firebase-backed apps from a prompt, with no local installation needed.

This is less of a shutdown and more of an upgrade. By folding Antigravity's agent harness directly into AI Studio, Google created a single surface for the entire journey from "I have an idea" to "here's a working app deployed to Firebase." The experience of building a Firebase app just changed significantly — in a good direction, once you've oriented yourself.

What Is AI Studio's Build Mode?

Google AI Studio was originally known as a prototyping playground for the Gemini API. In March 2026, it gained a major new layer: Build mode.

Build mode embeds the Antigravity coding agent directly into the AI Studio browser interface. From a single tab, you can:

• Describe what you want to build in plain language
• Watch the Antigravity agent generate and iterate on code
• Get Firebase (Firestore / Authentication) automatically provisioned as your backend
• Preview, test, and deploy — all without leaving the browser

The key difference from the standalone Antigravity desktop IDE: there's nothing to install. A Google account is all you need. Head to Google AI Studio and click "Build" in the sidebar to start.

What you're actually getting is a browser-resident version of the Antigravity agent — the same one that drives the desktop IDE — with Firebase provisioning wired in at the infrastructure level. When you describe an app that needs a database and authentication, the agent doesn't just write the code; it also sets up the Firebase project, configures Firestore, enables Authentication, and generates the environment variables. All of that used to require separate manual steps.

Step-by-Step: From Prompt to a Working App

Here's a concrete example — building a task manager with Google login using Build mode.

Step 1: Open Build Mode

In Google AI Studio, click "Build" in the left sidebar. On your first visit, you'll be prompted to create a project.

Step 2: Describe Your App

Build a simple task manager app.
Requirements:
- Google Sign-In via Firebase Authentication
- Tasks stored per user in Firestore
- Add, complete, and delete tasks
- Built with Next.js, TypeScript, and Tailwind CSS

The Antigravity agent parses this, plans the file structure, generates the code, and automatically provisions the Firebase resources. As it works, it surfaces "Artifacts" — a task list, implementation plan, and screenshots — so you always know what it's doing. This transparency is useful: you can intervene at any checkpoint, ask the agent to adjust something, or roll back a specific decision before it propagates through the codebase.

Step 3: Review the Auto-Generated Firebase Config

The generated project includes a properly structured Firebase initialization file:

// lib/firebase.ts (auto-generated)
import { initializeApp, getApps } from "firebase/app";
import { getAuth, GoogleAuthProvider } from "firebase/auth";
import { getFirestore } from "firebase/firestore";
 
const firebaseConfig = {
  // Populated from your auto-provisioned Firebase project
  apiKey: process.env.NEXT_PUBLIC_FIREBASE_API_KEY,
  authDomain: process.env.NEXT_PUBLIC_FIREBASE_AUTH_DOMAIN,
  projectId: process.env.NEXT_PUBLIC_FIREBASE_PROJECT_ID,
  storageBucket: process.env.NEXT_PUBLIC_FIREBASE_STORAGE_BUCKET,
};
 
const app = getApps().length === 0 ? initializeApp(firebaseConfig) : getApps()[0];
export const auth = getAuth(app);
export const db = getFirestore(app);
export const googleProvider = new GoogleAuthProvider();

API keys are stored as environment variables — never hardcoded. The agent handles the Firebase project linking automatically, which removes one of the most tedious parts of setting up a new Firebase app from scratch.

Step 4: Firestore Security Rules Are Generated Too

One detail that tends to get skipped in early prototypes: Firestore security rules. Build mode generates appropriate rules alongside your code:

// firestore.rules (auto-generated)
rules_version = '2';
service cloud.firestore {
  match /databases/{database}/documents {
    // Each user can only read and write their own tasks
    match /users/{userId}/tasks/{taskId} {
      allow read, write: if request.auth != null && request.auth.uid == userId;
    }
  }
}

"I forgot to set security rules and left everyone's data exposed" is a classic early-stage mistake. Having the agent generate sensible defaults closes that gap before you ever reach a real user. Once you're ready for production, you can refine these rules further — but starting from a secure baseline is the right call.

AI Studio Build Mode vs. Antigravity Desktop IDE

The reasonable question is whether the desktop Antigravity IDE is now redundant. In my experience, they're not competing — they cover different stages of a project's lifecycle.

AI Studio Build mode is the right choice when:

• You want to prototype a new idea fast, with a Firebase backend included
• You're doing a demo or early-stage exploration where speed matters more than depth
• You're working somewhere you don't have your development machine
• You're collaborating with non-technical people who want to follow along in real time

Desktop Antigravity IDE is the right choice when:

• You're adding features to an existing codebase with complex context
• You need local debugging, custom linters, or specific tooling that can't run in a browser
• You want to run multiple parallel agents using Manager Surface for larger tasks
• You need fine-grained context control via .antigravityignore and custom Rules

There's also a handoff that works surprisingly well: prototype in AI Studio Build mode until the core idea is validated, then export to a GitHub repository and continue in the Antigravity desktop IDE for production-grade development. The export is a standard repo — no proprietary format to deal with. This two-phase approach has meaningfully sped up my own workflow for new projects.

For a deeper look at what the fullstack vibe coding experience looks like end-to-end, the AI Studio 2.0 fullstack vibe coding guide is worth reading alongside this one.

Migrating a Firebase Studio Project

If you have an active Firebase Studio project, here's the migration path:

1. Get your code out

Export your project from Firebase Studio, or use the connected GitHub repository if you set one up. Firebase Studio had an export option that packages your project files as a ZIP or pushes to GitHub directly.

2. Open it in AI Studio Build mode

Choose "Open existing repository" and connect your GitHub repo. Before making any changes, the Antigravity agent reads your existing file structure and dependencies, and provides an analysis. This is a good moment to review what it found — sometimes it surfaces dependency conflicts or deprecated API usage worth knowing about.

3. Reconnect your Firebase project

Here's the important thing: your actual Firebase project doesn't need to migrate. Firestore data, Authentication settings, existing users — all of that lives in your Firebase project, not in the development environment. You're only changing where you write and run your code. In the environment variables panel, enter the credentials from your existing Firebase project, and the connection is re-established.

4. Test and deploy

After confirming the app runs correctly in the preview environment, deploy via Firebase App Hosting or your preferred platform. For step-by-step production deployment, the Firebase App Hosting deployment guide covers the full setup in detail.

The migration is genuinely less painful than it might sound. The Firebase project itself — your data, your users, your configuration — doesn't move at all. You're essentially swapping the IDE for a better one.

What Google I/O 2026 Will Add to This

Google I/O 2026 runs May 19–20, with a confirmed session: "Agent-first workflows from prompt to production." Firebase's positioning as an "agent-native platform" is expected to be formalized there.

From what's been previewed, the AI Studio → Antigravity → Firebase pipeline will deepen further: more automated provisioning, tighter agent integrations, and a cleaner path from browser prototype to production deployment on Google Cloud. Google Cloud Run integration and deeper Vertex AI connections have also been hinted at in session descriptions.

If you want to move quickly on whatever gets announced at I/O, getting comfortable with the current workflow now is the smart preparation. The mental model carries over directly — the tooling just gets more capable.

The First Step Worth Taking

If the Firebase Studio sunset left you unsure where to go, the simplest move is to open AI Studio's Build mode and build something small — a personal tool, a quick prototype, anything real. The Antigravity agent handles the heavy lifting, and you'll have a working Firebase app in under 30 minutes.

Once you have the mental model for how Build mode works, the migration from Firebase Studio stops feeling like a disruption and starts feeling like an upgrade you didn't ask for but are glad you got.