# Google Antigravity

> Google's agentic development platform — an agent-first IDE and Manager surface where multiple agents work across editor, terminal, and browser, on Gemini 3.

Google Antigravity is Google's agentic development platform: an agent-first IDE plus a Manager surface that oversees multiple agents across editor, terminal, and browser. It launched with Gemini 3 in November 2025; I/O 2026 added a desktop app, CLI, and SDK. Its free Individual plan is generally available; it replaced Gemini CLI's free tier in June 2026.

Website: https://antigravity.google

Google Antigravity is Google's **agentic development platform**, launched alongside Gemini 3 in November 2025. It pairs an agent-first IDE with a **Manager surface** — a mission-control view where you spawn, orchestrate, and observe multiple agents working asynchronously across different workspaces. Agents don't just edit code: they act **across the editor, the terminal, and a browser**, and report back with **Artifacts** — task lists, implementation plans, screenshots, and browser recordings you can verify instead of reading raw logs.

At I/O 2026 (May 19, 2026), Google expanded it from an IDE into a product family — **Antigravity 2.0** added a standalone desktop app, the **Antigravity CLI** for terminal-first work, and an SDK for custom agent behaviors. It is also the designated successor to [Gemini CLI](/tools/gemini-cli): as of June 18, 2026, Gemini CLI's free, Pro, and Ultra service stopped, and Google points those users to Antigravity CLI.

## Highlights

- **Manager surface** — run several agents in parallel on different tasks and supervise them from one board, rather than babysitting a single chat.
- **Browser-using agents** — agents can drive a browser to verify the UI they just built, attaching screenshots and recordings as evidence.
- **Artifacts over logs** — plans, task lists, and recordings designed for human verification of agent work.
- **Gemini models, plus others** — launched with Gemini 3 Pro; as of September 2026 the agent models include Gemini 3.8, 3.7, and 3.6 Flash, Gemini 3.1 Pro, Claude Sonnet and Opus 4.6, and gpt-oss-120b, and the agent harness is co-optimized with newer Gemini releases.
- **A whole product family** — IDE, desktop app, terminal CLI sharing the same Core Agent Engine, and a Python SDK (Apache-2.0).

## In an AI-assisted workflow

The IDE and desktop app install from the site; the CLI is one line:

```bash
# Antigravity CLI (terminal surface)
curl -fsSL https://antigravity.google/cli/install.sh | bash

# IDE / desktop app: download from https://antigravity.google/download
```

Coming from Gemini CLI, the CLI keeps the concepts you've built around — agent skills, hooks, subagents, and extensions (as plugins) — and can export terminal sessions to the Antigravity 2.0 desktop app, with settings synced between the two.

> [!TIP]
> Antigravity's distinctive bet is *verification*: agents produce screenshots and browser recordings of what they did. Lean into it — ask for browser-verified evidence on UI tasks instead of trusting the diff alone.

## Good to know

Antigravity runs on macOS, Windows, and Linux and is **proprietary** — free for individuals on a generally available plan, with paid Google AI subscriptions adding higher usage limits. The closed source is a real change from Gemini CLI's Apache-2.0 openness and has drawn criticism from contributors; weigh it if open tooling matters to your team ([OpenCode](/tools/opencode) and [Codex CLI](/tools/codex-cli) are the open terminal-agent alternatives). Expect pricing, limits, and the model lineup to keep shifting.

---

_Source: https://agentscamp.com/tools/antigravity — Tool on AgentsCamp._
