Google Antigravity Tutorial for Beginners: Build an App

Google Antigravity is Google’s agent-first code editor. You describe an app, an AI agent writes a plan, you review it, then the agent writes the code and tests it in a real Chrome window. In this Google Antigravity tutorial we install it, learn the Editor and the Agent Manager, and build a help desk ticket board in Next.js and Tailwind from one prompt. The free plan is enough to follow along, and Antigravity works in Saudi Arabia.

Watch the video (about 8 minutes):

How I checked this. Every fact here comes from Google’s Antigravity docs and FAQ, checked on 6 November 2026. Antigravity changes fast, so a button may sit in a slightly different place on your machine. The screens in my video are simplified look-alikes, not screen recordings.

What you need

• A computer with Windows, macOS or Linux.
• A personal Google account (Gmail). It works on the free plan.
• A little HTML or JavaScript. You don’t need to know Next.js; the agent writes it.
• Node.js installed, because the agent will run npx create-next-app for this demo.
• Google Chrome, for the browser agent that tests your app.

Step 1: Install Antigravity and sign in

1. Go to antigravity.google/download and pick the installer for Windows, macOS or Linux.
2. You’ll see more than one Antigravity:
3. Antigravity App: the agent-first desktop app.
4. Antigravity IDE: looks like VS Code, with agents built in. We use this one, because you can see every file the agent writes.
5. Antigravity CLI: agents in your terminal. There’s also a VS Code extension and a Python SDK.
6. Run the installer, open Antigravity, and sign in with your Google account.
7. On first launch you can import your VS Code settings or start fresh. Either is fine.
8. Open an empty folder. I called mine help-desk-board. An empty folder means the agent starts clean and doesn’t touch other projects.

Step 2: Editor, Agent Manager and the two modes

The IDE has two main windows:

• Editor: your files on the left (file tree, search, source control), the code in the middle, and the agent panel on the right, where you talk to the agent.
• Agent Manager: mission control. It lists every agent conversation in every workspace, and shows what each one is doing.

Switch between them with Ctrl+E on Windows and Linux, or Cmd+E on a Mac.

Before you send a task, pick a mode:

In newer versions you can also type /plan in the chat to start planning.

Then pick a model. On the free plan I use Gemini 3.8 Flash: it’s fast and it’s included.

Step 3: Write your first prompt (Planning mode)

Here’s the exact prompt from the video. Copy it into the agent panel with Planning mode selected:

Build a help desk ticket board with Next.js and Tailwind.
Three columns: Open, In progress, Closed.
Each ticket has a title, a priority and a requester.
Use sample tickets from Abd almouhimen Shamout.
Then run it and test it in the browser.

Notice what the prompt says: the stack, the columns and the data. A vague prompt gets a vague app. If you only type “make me a ticket app”, the agent has to guess the framework, the fields and the layout, and you’ll spend the next hour correcting it.

Step 4: Review the implementation plan

In Planning mode the agent doesn’t write code yet. It creates an implementation plan artifact with:

• Files it will create, for example app/page.tsx, components/Board.tsx, components/Column.tsx, components/TicketCard.tsx and data/tickets.ts.
• Steps in order.
• Verification: how it will check the result (here, open localhost:3000 in Chrome and move a ticket).

This is your moment. You can comment on any line, like a Google Doc. I added one comment: “Let me drag a ticket between columns.” The agent updated the plan with a new drag-and-drop step.

Fixing a plan takes seconds. Fixing finished code takes much longer. When the plan looks right, click Proceed.

Step 5: Watch the agent build and test

After you click Proceed:

1. A task list appears, and each item ticks off as the agent works.
2. The agent asks to run a terminal command, such as:

bash
npx create-next-app@latest . --ts --tailwind

Antigravity asks before it runs commands unless you change that in settings. As a beginner, keep it asking and read each command before you accept. If a command fails, the agent reads the error and tries a fix on its own.
3. In the editor you can watch the components appear: the board, the column and the ticket card.
4. The browser agent opens Chrome, loads localhost:3000 and clicks through the board to test it. The first time, it asks you to install the Antigravity Chrome extension. That’s expected.
5. When it’s done you get a walkthrough artifact: what changed, screenshots, and how to run it yourself:

bash
npm run dev
# then open http://localhost:3000

Want a change? Ask in the same conversation. I switched to Fast mode and asked for “red badges on high-priority tickets”. It changed one file in a few seconds.

My advice from client work: read the diff before you accept it. The agent is fast, but you own the code. If it goes to a client, you’re the one who answers for it.

Free vs paid: what you really get

Prices outside the US are different, so check the price in your own country on your Google subscription page. Model lists change often; Google’s docs say availability can change.

Arabic support and Saudi Arabia

• Saudi Arabia: supported. It’s on Google’s official list of countries where Antigravity is available.
• Syria: not supported yet. Users have asked Google to add it.
• Interface language: English. Google doesn’t list Arabic as an interface language.
• Arabic prompts: yes. You can write prompts in Arabic and the Gemini models understand them. For code I still write prompts in English, because the results are more predictable.

Errors and fixes

1. “Your current account is not eligible for Antigravity, because it is not currently available in your location”
Google checks the country on your Google account, not just your network. Use a personal Gmail account from a supported country, or request a region change from Google if the account’s country is wrong.

2. “Agent terminated due to error”
Retry once. If it happens again, start a new conversation or switch the model. If you added MCP servers, turn them all off, then bring them back one by one to find the one that breaks it.

3. Quota reached or 429 RESOURCE_EXHAUSTED
Wait for the quota to refresh (weekly on Free, every 5 hours on Pro), or switch to Gemini 3.8 Flash. Browser testing uses quota fastest, so test at the end, not after every small change.

4. The browser agent can’t open your page
Install Chrome, accept the Antigravity extension, and make sure your dev server is running on the right port (npm run dev, usually localhost:3000).

Checklist

• [ ] Installed Antigravity and signed in with a personal Google account
• [ ] Opened an empty project folder
• [ ] Chose Planning mode and a model
• [ ] Wrote a prompt with the stack, the features and the data
• [ ] Reviewed and commented on the implementation plan before clicking Proceed
• [ ] Read each terminal command before accepting it
• [ ] Read the diff and the walkthrough before shipping

FAQ

Is Google Antigravity free?
Yes. The free plan costs $0 and includes unlimited Tab completions and a weekly agent quota with Gemini 3.8 Flash and GPT-OSS-120b. Google AI Pro ($19.99 a month in the US) gives a bigger quota and more models.

Is Antigravity available in Saudi Arabia?
Yes. Saudi Arabia is on Google’s official list of supported countries. Your Google account’s country must be a supported one.

Does Antigravity support Arabic?
The interface is in English. You can write prompts in Arabic and the Gemini models understand them, but English prompts give more predictable code.

What’s the difference between Planning and Fast mode?
Planning mode writes a plan you can review and comment on before any code is written. Fast mode acts right away. Use Planning for multi-file work and Fast for small fixes.

Can I use Claude models in Antigravity?
Yes, on Google AI Pro and Ultra: Claude Sonnet 5.5 and Claude Opus 5.5 are in the model list. The free plan uses Gemini 3.8 Flash and GPT-OSS-120b.

Keep learning

• Kimi K3.1 Pro Review: Can It Replace Claude for Laravel Devs?
• Seedance 2.5 vs Kling 4.0 vs Gemini Omni: Who Wins?
• From the Claude course: Claude AI Tutorial for Beginners
• Next in AI Tools: Google Stitch Tutorial for Beginners (design an app UI with AI, then turn it into a Next.js page)

Watch the full video: https://youtu.be/9nfrHtcpDuQ

Sources

• Antigravity docs, Get started: https://antigravity.google/docs/getting-started
• Antigravity docs, Plans: https://antigravity.google/docs/plans/
• Antigravity docs, Choose a model: https://antigravity.google/docs/models/
• Antigravity FAQ (supported countries, account rules): https://antigravity.google/docs/faq/
• Antigravity blog, Gemini 4 Argon in Google Antigravity: https://antigravity.google/blog/gemini-4-argon-in-google-antigravity/
• Google Codelab, Getting Started with Google Antigravity: https://codelabs.developers.google.com/getting-started-google-antigravity
• Google AI Developers Forum, region error thread: https://discuss.ai.google.dev/t/your-current-account-is-not-eligible-for-antigravity-because-it-is-not-currently-available-in-your-location/136457
• Google AI Developers Forum, “Agent execution terminated due to error”: https://discuss.ai.google.dev/t/antigravity-agent-execution-terminated-due-to-error/181727
• Google AI Developers Forum, request to add Syria: https://discuss.ai.google.dev/t/request-to-add-syria-to-antigravitys-supported-regions-following-the-lifting-of-us-sanctions/182152
• Pricing round-ups (US prices): https://www.cloudzero.com/blog/google-antigravity-pricing/ and https://www.layer3labs.io/guides/google-antigravity-pricing