Where do I start if I am new to Antigravity?

Start with the Install guide to get the browser running, then follow the Getting Started page for a first-run walkthrough. Once the browser is installed and you have completed one agent run, the Concepts section of the docs centre will give you a clearer mental model of how the plan-execute-review loop works.

Where is the Antigravity API reference?

The API reference is on the Antigravity API page. It covers the public extension API, the tool-registration interface for custom agent tools, the MCP server integration surface, and the CLI scripting interface for running agents from the command line.

How do the docs relate to the official Antigravity documentation?

This reference site is independent of Google. The content here is researched from public sources including the official Antigravity developer documentation, release notes, and direct product testing. For authoritative SDK specifications and API contracts, always cross-check against the official Google documentation.

What category should I read if I am evaluating Antigravity for a security-sensitive environment?

Start with the Security category: the sandbox model page, the credential scopes documentation, and the audit logging reference. Then review the Login and identity reference for the Workspace tenant architecture. The release notes page is also relevant — it tracks changes to the security model across versions.

Antigravity docs centre — start here

Reading Guide

This hub links every reference page on the site by category. New users should read Install, then Getting Started. Developers extending the product should read Concepts, then API. Security and compliance teams should read Security first, then Login and identity. The release changelog is useful for everyone tracking what changed between versions.

The Antigravity reference project organizes its content into six categories. Each category is a collection of pages covering one aspect of the AI-native browser — its installation, its core concepts, its programmatic surface, its security model, its identity layer, and its release history. This page is the index for all of them. If you know what you are looking for, jump to the relevant section. If you are not sure, the navigation guide below each category description will point you in the right direction.

Concepts — how the browser works

The concepts category covers the foundational ideas behind the product: what a plan-execute-review loop is and why it matters, how the multi-window orchestration model differs from a standard tabbed browser, what the artifacts runtime captures and why, and how the Chromium engine and the agent runtime divide responsibilities. These pages are not tutorials; they are mental-model builders. Reading them before diving into the hands-on guides will make the hands-on material faster to absorb.

Key pages in this category: the main reference page provides a concise product overview. The features page catalogues every major capability. The AI agent page explains the planning and execution model in depth.

Install — getting the browser running

The install category covers the full setup process across platforms. There are dedicated pages for Windows, macOS, and Linux, plus a guide to the Chrome extension bridge for teams that need to keep existing Chrome profiles working alongside the new tool. System requirements, download verification, first-launch setup, and Google account sign-in are all covered. This is where new users should start.

Key pages: install guide (main), Windows download, download page, getting started (first-run walkthrough after install).

API — extending and scripting the browser

The API category documents the programmatic surface that developers use to extend the agent runtime. This includes the tool-registration interface for adding custom tools to the agent's planning vocabulary, the MCP server integration model, the CLI scripting interface for running agents from automation pipelines, and the extension API for browser-level integrations. The target audience for this section is developers building on top of the product rather than just using it.

Key pages: Antigravity API reference, developer tools. The OECD AI Policy Observatory has background reading on the policy considerations around programmable AI agents if that context is useful for your evaluation.

Security — isolation, credentials, and audit

The security category covers the sandbox architecture, the credential scope model, the artifact storage boundary, the offline-first operating mode, and the audit logging subsystem. This is the section for security engineers evaluating the product and compliance teams assessing it for regulated deployment. The pages in this category are kept especially current — any change to the security model in a new release triggers an immediate review of the relevant pages.

Key pages: security and sandbox model, login and identity reference.

Agents — the runtime in depth

The agents category sits between concepts and API. It covers the agent runtime from a practitioner's perspective: how task planning works in detail, how checkpoints are configured, how the agent handles errors and uncertain states, how multi-window orchestration is set up, and how to read and act on the artifact timeline after a run completes. This is the section for engineers who are past the getting-started phase and want to use the tool more deliberately.

Key pages: AI agent deep dive, features overview. The Stanford HAI research hub has published work on human-in-the-loop agent design that provides useful theoretical grounding for the checkpoint model.

Changelog — what changed and when

The changelog category tracks every public Antigravity release: version number, release date, what changed, and any breaking changes relevant to developers using the API or extension surface. This is the section to check after a new release ships to understand whether anything in your workflow needs to change. The release notes page is updated within 24 hours of a new release going public.

Key pages: release notes and changelog.

Category	Pages	Who it's for
Concepts	Product overview, features, AI agent model	Anyone evaluating or starting with the browser
Install	Install guide, Windows/macOS/Linux, Chrome bridge	New users setting up the product
API	API reference, tool registration, CLI scripting	Developers extending or automating the browser
Security	Sandbox model, credentials, audit log, offline-first	Security engineers, compliance teams
Agents	Runtime deep dive, checkpoints, artifacts	Power users and teams running complex tasks
Changelog	Release notes, version history	Everyone tracking product changes

Docs centre — four questions new readers ask

Common orientation questions from readers arriving at the docs centre for the first time.

Where do I start if I am new to Antigravity?

Start with the install guide to get the browser running on your machine, then follow the getting started page for a first-run walkthrough that covers signing in, creating your first task, and reading the artifact timeline after it completes. Once you have done one run, the concepts pages will make much more sense because the product makes most sense as a felt experience rather than a description. Budget about thirty minutes for the full new-user sequence.
Where is the Antigravity API reference?

The API reference lives on the Antigravity API page. It covers the public extension API, tool-registration interface, MCP server integration, and CLI scripting surface. The developer tools page covers the browser DevTools extensions and the artifact inspection interface. If you are building a custom tool that the agent can use during a run, the tool-registration section of the API page is where to start.
How do the docs relate to the official Antigravity documentation?

This reference site is independent. It is not affiliated with Google and does not reproduce proprietary content. The content here is researched from public sources — official release notes, the official developer documentation, and direct product testing. For authoritative API contracts, SDK specifications, and support, use the official Google documentation and support channels. This site's value is in synthesis, context, and cross-referencing; it is not a substitute for the official source on technical specifics.
What category should I read if I am evaluating the browser for a security-sensitive environment?

Start with the security and sandbox model page, which covers per-run isolation, credential scopes, the artifact boundary, offline-first mode, and the audit log in detail. Then read the login and identity reference for the Workspace tenant architecture and the identity boundary between personal and work accounts. The release notes page is also worth reviewing — it tracks changes to the security model across versions, which matters if you are assessing a specific version against a security baseline.