What if your code editor could not only suggest code, but code for you — plan, execute, test, and verify, all autonomously? That is the promise of Google Antigravity, the agentic development platform launched by Google in November 2025. Built on a fork of VS Code after the acquisition of the Windsurf team for $2.4 billion, Antigravity marks a turning point: we are moving from code assistance to agent orchestration.
From Windsurf to Antigravity: the origin story
In July 2025, Google makes a big bet: the acquisition of the Windsurf team for $2.4 billion. Windsurf was already known for its agentic approach to development. In just four months, Google transforms that expertise into Antigravity — launched on November 18, 2025 alongside Gemini 3. The speed of execution is remarkable.
How it works: Editor + Manager
Antigravity introduces a dual-view architecture that changes the game:
- Editor View — The classic coding environment. Tab autocompletion, natural language commands, compatible VS Code extensions. Your usual workflow, but better.
- Manager View — The real differentiator. A dedicated interface where you launch, orchestrate, and supervise multiple AI agents in parallel. Each agent can code, use the terminal, browse the web, and generate verifiable artifacts of its work.
The philosophy is clear: AI should no longer be a chatbot in the sidebar, but an autonomous actor with its own workspace. You provide a high-level task, the agent breaks it down, executes it, and delivers verifiable results.
Gemini 3.1 Pro under the hood
Antigravity is powered by [Gemini 3.1 Pro](/blog/gemini-3-1-pro-google-deepmind-reasoning), the latest model from Google DeepMind. With a 2 million token context window, agents can index and reason over entire codebases — not just individual files. This is a major advantage for large-scale projects.
The platform also supports other models: Claude Sonnet 4.6 from Anthropic and GPT-OSS from OpenAI. This model optionality is an asset compared to more locked-down competitors.
Antigravity vs the competition
| Criterion | Google Antigravity | Cursor | Claude Code | GitHub Copilot |
|---|---|---|---|---|
| Approach | Agent-first, dual-view | AI integrated in editor | Agentic terminal | Autocompletion + chat |
| Base | VS Code fork | VS Code fork | CLI | VS Code extension |
| Main model | Gemini 3.1 Pro | Multi-model | Claude Opus 4.7 | GPT-5.4 |
| Context | 2M tokens | Variable | 1M tokens | Variable |
| Parallel agents | Yes (Manager View) | No | Yes (background tasks) | No |
| Built-in browser | Yes | No | No (Computer Use) | No |
| Price | Free (preview) | $20/month | Included in Max | $10/month |
| Verifiable artifacts | Yes | No | No | No |
AI code tools comparison (April 2026)
Limitations in April 2026
Antigravity is not without flaws. The platform is still in preview and has notable stability issues. Agents can sometimes produce inconsistent results on complex tasks, and the Manager View still lacks maturity for production use. With roughly 6% adoption among developers after two months, it is promising but far from a standard.
Who is Antigravity for?
Antigravity targets developers who want to move from assisted coding to orchestrated coding. If you spend your days juggling between the editor, terminal, and browser to test your changes, Antigravity proposes delegating that entire loop to an agent. It is also an excellent entry point into the Google Cloud ecosystem, with native integration for serverless deployment.
Compare AI coding tools
Frequently asked questions
Stay up to date on AI news
Get our analyses of the latest advances in artificial intelligence.
No spam. Unsubscribe in 1 click.
Gemini 3.1 Pro: the engine behind Antigravity
Discover the model powering Google's agentic platform.
Read the articleAI Model Comparison Guide Q2 2026
GPT-5.4, Opus 4.7, Gemini 3.1 Pro — which model should you choose?
Read the guide

