🚀 For teen builders

Google AI Studio
Your lab is open.

Build real full-stack apps with AI. Free, in your browser. No coding experience required to start.

You type →
Free to use
No installation
Any Google account
Open AI Studio →
Start here

Not the chatbot you know

Google has two separate AI products. Most people mix them up. Here is the difference.

💬
Gemini App
The consumer chatbot. You ask questions and it answers. The model, the settings, and the tools are all hidden. Like texting a smart friend.
🧪
Google AI Studio
The workshop. You control the model, the settings, and the tools. You test prompts, build real full-stack apps, and export actual code. You run the lab.
Gemini app = watching Netflix  |  AI Studio = operating the film studio
Inside AI Studio

Three modes, three jobs

When you open AI Studio, you choose a mode. Each one does something different.

💬
Playground
Your testing ground. Chat with any model, change settings like Temperature or System Instructions, and try the same prompt on multiple models at once using Compare Mode. This is where you learn what the AI can and cannot do before you build.
✅ Best for: experimenting with prompts
🔨
Build mode
The main event. You describe an app in plain English and the Antigravity agent generates a working full-stack app — frontend, backend, and database together. A live preview appears on the right side of your screen. You keep asking for changes, the agent keeps updating everything. This is vibe coding.
✅ Best for: building your first full-stack app
🎙️
Stream / Live
Talk to the AI in real time using your voice. You can also share your screen and the AI can see exactly what you are working on and help you live. Think of it as a video call where your AI tutor can see your screen and respond out loud.
✅ Best for: live help while you work
Pick the right brain

The models, explained simply

A model is the AI engine behind every answer. All five are free to test inside AI Studio.

⚡
Gemini Flash
Fast. The default. Handles most everyday tasks well at high speed. Start here for most builds.
Best for: most tasks
🧠
Gemini Pro
Slower but deeper. Complex reasoning, long analysis, and hard multi-step problems.
Best for: hard questions
🍌
Nano Banana
Image generator. Creates 2K and 4K images. One of the few that reads text inside images correctly.
Best for: making images
🎬
Veo
Text-to-video. Describe a scene and it generates a clip with motion and transitions.
Best for: video clips
🎵
Lyria
Music and audio generator. Specify a genre, mood, and tempo, and it produces original sound.
Best for: music and audio
Full stack simplified

Every app has three layers

Most AI tools only build one layer: the front. AI Studio builds all three automatically with one prompt.

🖥️
Layer 1
Frontend
React
The screens, buttons, and everything users see and click on.
→
⚙️
Layer 2
Backend
Node.js
The server-side logic that runs when a user takes an action.
→
🗄️
Layer 3
Database
Firestore
Where your data lives after the app closes. Permanent memory.
The Antigravity agent reads your prompt and writes all three layers at the same time, making sure they all connect correctly.

Real-world analogy: a food app

📱
Menu screen
Frontend (React)
👨‍🍳
Kitchen logic
Backend (Node.js)
🗂️
Order history
Database (Firestore)
🚪
Login screen
Auth (Firebase)
What you get

The full stack the agent generates

When you describe an app in Build mode, the Antigravity agent provisions all six of these — no configuration needed from you.

1
Frontend
React (default) · Angular · Next.js
The full user interface: all screens, buttons, forms, and navigation. Specify "build this with Next.js" in your prompt to switch framework.
2
Backend
Node.js runtime (server-side)
Handles API routes, business logic, secure calls to outside services, and npm package usage. The agent installs packages for you when you ask.
3
Database
Firebase Firestore (NoSQL)
Auto-provisioned cloud database. Stores and syncs data in real time. No SQL required. Supports offline capability and multi-device sync out of the box.
4
Authentication
Firebase Authentication
Secure Google Sign-in for your users. Just ask the agent to "add user login" and it writes and configures the whole auth flow automatically.
5
Hosting
Google Cloud Run
One-click publish gives you a live .run.app URL in seconds. HTTPS-secured and scalable. Real-world traffic ready from day one.
6
AI features
Gemini API (server-side, auto-secured)
Any AI feature in your app calls Gemini from the server side. Your API key is auto-configured as a secret — never exposed in the browser.
Bonus integrations the agent handles automatically: Google Workspace (Gmail, Sheets, Drive, Calendar), third-party REST APIs, npm packages from the full npm ecosystem. Ask in plain English and the agent wires it up.
Technical specifications

The numbers you need to know

Real specs, no fluff. These are the limits and capabilities as of June 2026.

1M
token context window
(Gemini 3.5 Flash)
65K
max tokens per
single response
3
web frameworks:
React, Angular, Next.js
6
full stack layers
built automatically
1
click to publish
a live URL
Free
to prototype and
test in AI Studio
Layer Technology What it does Free?
FrontendReact / Angular / Next.jsAll screens and UI components✅ Free
BackendNode.js runtimeServer logic and API routes✅ Free
DatabaseFirebase FirestoreNoSQL cloud database, real-time sync⚡ Spark plan free
AuthFirebase AuthenticationGoogle Sign-in for your users⚡ Spark plan free
HostingGoogle Cloud RunLive .run.app URL, HTTPS, scalable⚡ Free for prototyping
Mobile (Android)Kotlin + Jetpack ComposeNative Android app in browser emulator✅ Free
AI (in-app)Gemini APIPowers any AI features you add⚡ Free tier, then $1.50/M tokens
Free tier context: AI Studio itself is always free. Firebase's Spark plan covers typical prototype workloads for Firestore and Auth. Cloud Run gives you a free quota for deployment. You only pay if your app gets real traffic and exceeds the free quotas — which will not happen while you are learning.
Power-ups

Four built-in tools

These are toggles in the settings panel. Turn them on to give the model extra abilities.

🔍
Search grounding
The AI searches Google before answering. Turn on for current events, prices, or recent news where you need facts from today, not last year.
🗺️
Maps grounding
Connects the AI to real map data so it can verify addresses, distances, and location info. Note: Search and Maps grounding are alternatives — not both at once.
💻
Code execution
The AI can write and run Python code for you — useful for maths, charts, or processing data. It shows you the result, not just the code.
🔌
Function calling
Advanced. Lets the AI connect to and trigger actions in outside apps and services. This is how you build agents that do real things, not just talk about them.
Full stack how-to

Build a real app, step by step

This is the full flow from blank prompt to live URL. Use this exact sequence every time you build in AI Studio.

Example full-stack prompt to copy
"Build a book reading tracker. Users sign in with Google. They can add books with a title, author, and finish date, mark each book as reading or done, and see a count of books they finished this year. Store all data so it saves between sessions."
This prompt tells the agent: who logs in, what data is stored, what screens exist, and that persistence is needed. The agent will auto-detect it needs Firebase and prompt you to enable it.
1
Open aistudio.google.com and sign in
Any Google account works. No setup, no download, no card. Click Build in the left sidebar. You will see a prompt box on the left and a live preview panel on the right.
2
Write a specific full-stack prompt
Name the users, the data they store, the screens they see, and whether the data should persist. The agent reads all of this and plans the complete frontend, backend, and database structure before writing a single file.
⚡ Say "store all data" or "save between sessions" to trigger Firebase auto-provisioning
3
Enable Firebase when the settings card appears
If your app needs a database or login, the agent automatically shows a Firebase configuration card. Click the settings icon, pick a region (this also sets your Cloud Run region), then confirm. The agent then auto-provisions Firestore and Authentication and writes all the integration code. You touch nothing.
⚡ Pick a region close to your users — Southeast Asia = asia-southeast1
4
Watch the agent work — then test the preview
The Antigravity agent plans the project, writes all files across all layers, runs the app in a built-in browser to check for errors, and fixes what breaks — all before showing you the result. When the preview appears, click through the app and test it like a real user.
5
Ask for one change at a time
Do not send five requests at once. Pick one thing: "Make the add-book form slide in from the right" or "Change the done books to show in green." The agent updates all affected files together, then you test again. Use the Annotate button to draw directly on the preview and point to what needs changing.
⚡ Approve agent checkpoints — it pauses before big structural changes
6
Publish to get your live URL
Click Publish → Get Started → Publish App. The agent packages everything and provisions a Cloud Run service. Within seconds you have a live .run.app URL anyone can open. Share it, test it on your phone, or add it to a portfolio.
⚡ On the free tier, keep real user data out — your prompts may train Google's models
🌐
Cloud Run
Publish → live .run.app URL in seconds. Share with anyone.
📦
GitHub export
Push the full codebase to a repo for real production development.
💾
ZIP download
Download all files and run locally with npm install then npm run dev.
Terminology

15 words you need to know

Tap any card to see what the term means and an analogy to make it stick. Search to find a specific one fast.

🔎

Showing all 15 terms

Test yourself

Quick knowledge check

Five questions. See how much landed.

Question 1 of 5

Build something today 🚀

Pick one idea. Open Build mode, write your prompt, and get to a live URL before this session ends.

🎯
Quiz app
5 questions, 4 options, score at end. Add a leaderboard that saves scores with Firebase.
🍜
Recipe finder
User types ingredients, app suggests recipes. Let users save favourites to Firestore.
⏱️
Study timer
25-minute focus countdown. Log sessions and show a weekly history chart.
🌤️
Mood tracker
Log mood daily, store per user in Firestore, show a chart of the week.
Start building →