How to Use Google Antigravity IDE: AI Agents Build Apps

What is Google Antigravity?

Google Antigravity is a free AI-powered code editor (IDE) where AI agents write, test, and debug code for you. Think of it as having a team of AI developers working on your project at the same time.

Who is this for? Beginners, non-coders, and developers who want to ship faster.

What you'll learn: Install Antigravity, configure agents, and build a working app using plain English instructions.

Time needed: ~15 minutes

---

Prerequisites

• A personal Gmail account (Google Workspace accounts not supported yet)
• Windows, macOS, or Linux
• Chrome browser installed

---

Step 1: Download and Install

1. Go to antigravity.google/download
2. Select your OS (Mac .dmg / Windows .exe / Linux .deb)
3. Run the installer
4. Launch Antigravity

---

Step 2: Sign In and Initial Setup

1. Sign in with your personal Gmail account
2. Choose a theme (Dark/Light) — or import settings from VS Code or Cursor
3. Configure two key settings:

Setting	Option	What It Does
Terminal Permissions	"Request Review"	Agent asks before running commands
Artifact Review Policy	"Pause on Major Changes"	Agent pauses before big file edits

Recommended for beginners: Keep both on "Request Review" so you stay in control.

---

Step 3: Understand the Two Views

Antigravity has two main interfaces:

Editor View

Your standard code editor. Write code, get autocomplete suggestions, and chat with AI in the sidebar.

Manager View (Mission Control)

This is where the magic happens. You type a task in plain English, and Antigravity assigns AI agents to complete it.

How to switch: Click the toggle icon in the top-right corner of the window.

---

Step 4: Run Your First Mission

1. Open Manager View
2. Click "New Mission"
3. Type your task in plain English

Example Input:

Build a landing page with a hero section, feature cards, 
and a contact form. Use HTML, CSS, and JavaScript.

1. Click "Run"
2. Watch the agents work in real time

Example Output: Antigravity creates 3 files:

index.html    — Page structure with hero, cards, contact form
styles.css    — Responsive styling
script.js     — Form validation logic

Each agent shows its progress as visual artifacts (plans, code diffs, screenshots).

---

Step 5: Review and Approve Changes

Agents pause at checkpoints and show you:

• Task Plan — What the agent intends to do
• Code Diffs — What lines will change
• Browser Screenshots — How the result looks

Click "Approve" to continue or "Edit" to give corrections.

Example correction:

Change the hero background color to dark blue (#1a1a2e) 
and make the CTA button larger.

The agent updates only what you asked — no full rebuild.

---

Step 6: Use Multiple Agents in Parallel

This is Antigravity's biggest advantage. You can run up to 5 agents simultaneously.

Example Mission:

Agent 1: Build user authentication with Google OAuth
Agent 2: Write unit tests for the auth module
Agent 3: Create a dashboard page with charts
Agent 4: Set up a SQLite database for user profiles
Agent 5: Write API documentation in Markdown

Each agent works independently and reports back when done.

---

Step 7: Choose Your AI Model

Antigravity supports multiple AI models. Assign different models to different agents based on the task.

Model	Best For
Gemini 3.1 Pro	General coding, planning
Gemini 3 Flash	Quick edits, fast iteration
Claude Sonnet 4.6	Nuanced logic, debugging
Claude Opus 4.6	Complex architecture

How to change: Click the model name next to any agent in Manager View and select from the dropdown.

---

Step 8: Test with the Built-In Browser

Antigravity has a built-in Chrome-based browser. Agents automatically:

1. Launch your app in the browser
2. Take screenshots of the result
3. Report visual bugs they find

You don't need to manually open a browser to test.

---

Pricing Quick Reference

Plan	Cost	Requests/Day
Free	$0	~20
Pro	$20/month	Higher limits
Ultra	$249.99/month	Maximum access

The free tier is enough to try everything in this guide.

---

Common Mistakes to Avoid

1. Vague missions — "Make a website" is too broad. Be specific about pages, features, and tech stack.
2. Skipping reviews — Always check agent output before approving. Agents are good but not perfect.
3. Too many agents on small tasks — One agent is fine for a single-page component. Use multiple agents for multi-feature projects.

---

Quick Recap

Step	Action
1	Download from antigravity.google/download
2	Sign in with personal Gmail, configure review settings
3	Learn Editor View (code) vs Manager View (agents)
4	Type a mission in plain English and hit Run
5	Review agent output at each checkpoint
6	Run up to 5 agents in parallel for bigger projects
7	Pick the right AI model per task
8	Use the built-in browser for testing

---

Next Steps

• Try building a todo app with authentication as your first real project
• Explore AgentKit 2.0's 16 specialized agents for frontend, backend, and testing
• Connect your GitHub repo directly to push code from Antigravity