Google AI Studio Gets Full-Stack Vibe Coding, Powered by Antigravity

Google AI Studio just got a serious upgrade. The Build tab, which launched with basic vibe coding support back in October 2025, now handles full-stack applications with databases, authentication, multiplayer state, and a Node.js backend runtime. The Antigravity coding agent powers the whole thing.

Open post on X

Logan Kilpatrick, Product Lead for Google AI Studio, announced the update on March 19, 2026. The official blog post from Google (authored by Ammaar Reshi) lays out what changed.

What Actually Shipped

The short version: AI Studio’s Build mode went from a frontend prototyping tool to something that can produce apps with real backends, real databases, and real user authentication. Here’s what’s new.

One-click Firebase integration. Click a button and AI Studio auto-provisions a Cloud Firestore database and Firebase Authentication for your project. No console hopping, no config files, no OAuth setup. The Antigravity agent detects when your app needs persistent data or user accounts and can suggest enabling these services automatically.

Sign in with Google. Firebase Auth comes pre-wired with Google OAuth 2.0 support. Users of your app can sign in with their Google account out of the box. For a prototyping tool, this removes one of the most tedious parts of building anything user-facing.

Full Node.js backend runtime. This is the biggest functional change. Previously, Build mode was essentially a frontend sandbox. Now you get a server-side Node.js environment with access to the npm ecosystem. You can write API routes, handle server logic, and keep secrets out of client-side code.

Secrets management. Server-side environment variables, stored securely and only accessible from backend code. The Antigravity agent is aware of them and can auto-detect when your app references API keys that should be stored as secrets rather than hardcoded.

Multiplayer and real-time state. The backend runtime supports server-managed state synchronization. Google specifically calls out live chat apps, collaborative whiteboards, and multiplayer games as use cases. You can test multiplayer behavior by opening multiple browser tabs in Build mode, or by sharing the preview URL with someone else.

Third-party integrations. Stripe for payments, SendGrid for email, Supabase, MongoDB Atlas, or any REST API. The backend runtime can make outbound HTTP requests, which means your AI Studio app isn’t locked into Google’s ecosystem for everything.

The Antigravity Agent

The coding agent behind Build mode is Antigravity, the same technology that powers Google’s standalone desktop IDE. In the AI Studio context, the agent writes code, runs it, tests it, catches errors, and iterates until the app works. It maintains awareness of the entire project, manages dependencies across files, and handles the feedback loop between frontend and backend code.

Three ways to direct it: chat-based prompting (describe what you want), annotation mode (highlight a specific UI element and describe the change you want), or direct code editing in the built-in editor.

The underlying model is Gemini, with the agent architecture handling the planning and execution layers on top. If you’ve used Antigravity as a standalone IDE, the coding behavior should feel familiar. The difference is that AI Studio wraps it in a zero-setup browser environment.

Frameworks and Deployment

Build mode supports React (the default), Angular, and Next.js. Next.js support was added in February 2026.

For getting your app out of Build mode, there are three options: download as a ZIP, push to GitHub, or deploy directly to Google Cloud Run. The Cloud Run path is the simplest route to a live URL, though it incurs standard Google Cloud pricing. The GitHub export is probably the most practical for anyone who wants to keep iterating outside of AI Studio.

How This Fits Into Google’s Tool Lineup

Google now has three overlapping developer tools, and the boundaries between them are worth understanding.

Google AI Studio is the AI experimentation and prototyping environment. The Build tab is one part of a larger product that also handles model testing, prompt engineering, and API key management. Vibe coding is a feature here, not the whole product.

Firebase Studio (formerly Project IDX) is a full cloud-based development environment aimed at production-quality apps. It supports Flutter, Angular, Go, and other frameworks, with tighter integration into Google Cloud infrastructure. Think of it as Google’s answer to Replit or Codespaces, with Firebase services baked in.

Google Antigravity is the standalone desktop IDE, built on a VS Code fork, with the agent architecture as its core differentiator. It can run multiple autonomous agents simultaneously across different tasks, which is something the AI Studio integration doesn’t expose.

The new update draws a clearer line between these products: AI Studio handles “describe what you want and watch it get built,” Firebase Studio is for developers who want a full cloud IDE with manual control, and Antigravity is for developers who want agent-first workflows on their local machine. All three share Firebase backend services. Google’s own documentation acknowledges the overlap can be confusing.

Pricing

AI Studio’s Build mode is free to start, no credit card required. The Firebase integration (Firestore, Auth) is covered under Firebase’s free tier for prototyping-scale usage. Cloud Run deployment costs money at Google Cloud rates. Heavy API usage beyond the free tier follows Gemini API pricing.

For most people building prototypes or small apps, the free tier should be sufficient. Production-scale apps with meaningful traffic will hit paid thresholds, but that’s true of every platform in this category.

What This Means for the Builder Market

The app builder space has gotten crowded. Bolt, Lovable, Replit, and v0 all occupy some version of “describe an app and get working code.” Google’s entry has two structural advantages: it’s free to start (no separate subscription), and the Firebase integration gives you production-grade infrastructure (Firestore, Auth, Cloud Run) without leaving the platform.

The disadvantage is complexity. Google has three products that can all build apps, and the documentation doesn’t always make it obvious which one you should use for what. That’s a discovery problem, not a capability problem, but it matters when someone new to the ecosystem is trying to pick a tool.

The Antigravity agent is the technical differentiator worth watching. It’s the same agent architecture used in a standalone IDE that supports multi-agent workflows, now accessible in a browser with zero setup. If Google continues improving the agent’s capabilities at the pace they’ve shown since Antigravity’s November 2025 launch, the gap between “prototype” and “production app” in AI Studio will keep shrinking.

For now, if you want to test the new Build mode: go to aistudio.google.com/vibe-code, describe an app, and see what the Antigravity agent produces. The database and auth integrations are available immediately.

Sources: