Evergreen Tools
|
← Back to AI Tools

Google Antigravity Agent-First Development Platform

Instead of bolting AI into an editor sidebar, Antigravity makes the editor, terminal and browser all work surfaces for agents: the platform becomes a command center for multiple parallel local agents, free for individuals during public preview

Tool Interface

Interactive tool will be available soon

Features

  • ✓ Agent-first, not editor-first: the IDE is one view among several while the platform itself orchestrates agents that plan, execute and verify tasks
  • ✓ Command center for parallel agents: group conversations into projects, operate across multiple workspaces and automate routine work with scheduled messages
  • ✓ Action across surfaces: agents can work beyond the editor, and browser control lets them open pages, click and read results on their own
  • ✓ Four entry points: IDE, CLI, SDK and the manager, with a terminal surface for autonomous coding agents and background subagents plus a Python SDK for prototypes
  • ✓ Verifiable artefacts: tasks leave inspectable artefacts and activity records for human review instead of a black-box answer; public preview is free for individuals with generous Gemini 3 Pro limits

How to Use

  1. Download the installer for your platform from the official site (macOS, Windows and Linux are supported), sign in with a Google account and accept the defaults
  2. Start on a small project: give it one small end-to-end requirement and watch how it decomposes the task, which files it really touches and whether its verification is credible
  3. Group agents by project, define the repository boundary and allowed operations for each project, and attach scheduled messages for routine automation
  4. Keep a human gate: route important changes through code review and tests, treat agent output as a draft for review, and never hand production credentials to the agent

FAQ

What is Google Antigravity?

It is Google's agent-first development platform. The official launch post calls it our new agentic development platform: the core is a familiar AI-powered IDE experience, but it evolves the IDE toward an agent-first future with browser control and asynchronous interaction patterns. It was announced on 18 November 2025 alongside Gemini 3, and is free for individuals during public preview. See https://antigravity.google/blog/introducing-google-antigravity and https://developers.googleblog.com/build-with-google-antigravity-our-new-agentic-development-platform

How is it different from an ordinary AI editor?

It comes down to who is driving. Conventional AI editors embed completion and chat inside the editor while the human stays the executor. Antigravity states that its editor view is only the starting point and that the platform is moving toward an agent-first future with browser control and asynchronous interaction, letting agents autonomously plan and execute complex end-to-end software tasks. The editor, terminal and browser all become agent work surfaces. See https://antigravity.google/blog/introducing-google-antigravity

Does it cost anything yet, and which models are supported?

The official developer blog says it is available today in public preview, at no cost for individuals, with generous rate limits on Gemini 3 Pro. It also offers model optionality with full support for Anthropic's Claude Sonnet 4.5 and OpenAI's GPT-OSS, and it is cross-platform across macOS, Windows and Linux. Check current limits and model lists at https://antigravity.google/ and https://developers.googleblog.com/build-with-google-antigravity-our-new-agentic-development-platform

What are the CLI and SDK for?

They serve different habits. The official product page describes the CLI as a lightweight, fast, terminal-first surface for running autonomous coding agents, executing shell commands and managing background subagents, which suits keyboard-driven developers. The SDK lets you prototype custom agents on Antigravity's harness with minimal code — the official example is a simple Python script for iterating on agentic applications, automating software engineering tasks and running evaluations. See https://antigravity.google/ and https://codelabs.developers.google.com/getting-started-google-antigravity

How do you control the risk of letting it change code on its own?

Three moves. Go small first: run a small end-to-end task in a non-critical repository to learn its behavioural boundaries. Set boundaries and group work by project so you control which workspace and directories it may touch. Keep the human gate: route important changes through code review and tests, treat its output as a draft, keep production credentials out of its hands and retain auditable activity records. Google stresses building for user trust, but your own process still provides the guardrails. See https://antigravity.google/