AI Coding Tools
Google Antigravity Explained: Gemini's Agent-First Coding Tool
If you’ve searched “Gemini coding tool” recently and landed on a name you didn’t expect, that’s not a mistake. Google Antigravity is now the center of its developer AI lineup, and it quietly absorbed a tool a lot of people were still using: Gemini CLI.
Antigravity isn’t a chatbot bolted onto an editor. It’s a full rethink of what an IDE is for, built around the idea that you shouldn’t be typing every line yourself anymore. Here’s what it is, how it got here, and whether it’s worth your time.
From Gemini CLI to Antigravity: A Fast Timeline
Google announced Antigravity on November 18, 2025, alongside the Gemini 3 launch. At first it sat next to Gemini CLI as a separate, more ambitious product: an agent-first IDE rather than a terminal tool.
That didn’t last long. At Google I/O on May 19, 2026, Google announced it was folding Gemini CLI into a new Antigravity CLI, giving developers a 30-day migration window. Gemini CLI officially stopped working for Google AI Pro, Ultra, and free-tier users on June 18, 2026. If your org uses Gemini Code Assist Standard or Enterprise, or Code Assist for GitHub through Google Cloud, you were unaffected and kept your existing access.
The message was clear: Antigravity, not a standalone CLI, is where Google wants developers working now.
What Antigravity Actually Is
Antigravity is an agent-first IDE, built as a modified fork of VS Code, that shifts your job from writing code line by line to directing autonomous agents that plan, execute, and verify their own work.
The centerpiece is the Agent Manager, a view for orchestrating multiple agents running in parallel across different workspaces. Instead of babysitting one Copilot-style suggestion at a time, you hand off a task, an agent breaks it into steps, and you check back on progress like you would with a real teammate’s pull request.
The other standout feature is built-in browser control. An Antigravity agent can spin up a real Chromium instance, navigate to your local dev server, click through the user flow it just built, and capture screenshots as proof the feature actually works, not just that the code compiles.
By mid-2026, Antigravity had grown into four surfaces:
- Antigravity IDE — the VS Code fork, for hands-on editing alongside agents
- Antigravity 2.0 — a standalone desktop app
- Antigravity CLI — a Go-based terminal tool (Gemini CLI’s replacement) that runs agents in the background without locking up your session
- Antigravity SDK — a Python SDK for building your own agent workflows on top of the platform
Antigravity CLI kept the features developers relied on in Gemini CLI, including Agent Skills, Hooks, Subagents, and Extensions, just repackaged as Antigravity plugins.
Which Models Power It
Antigravity isn’t locked to Google’s own models. It supports Gemini 3 Pro, Gemini 3 Deep Think, and Gemini 3 Flash, plus third-party options including Claude Sonnet 4.6, Claude Opus 4.6, and GPT-OSS 120B. You pick the model per agent, so a quick refactor can run on Flash while a gnarly multi-file bug goes to Gemini 3 Pro or Claude Opus.
On SWE-bench Verified, a standard benchmark for real-world coding tasks pulled from actual GitHub issues, Antigravity has been reported scoring around 76.2%, competitive with the other top agentic coding tools on the market in 2026.
What It Costs
- 🆓 Individual / Free ($0/month): Basic weekly rate limits, unlimited tab completions, unlimited command requests, and access to the listed Google, Anthropic, and OpenAI models.
- 💰 Google AI Pro ($20/month): Higher usage limits across Antigravity’s surfaces, bundled with the rest of the Google AI Pro subscription.
- 🚀 Google AI Ultra ($100/month): Introduced for Antigravity on May 19, 2026, with the highest limits for developers running agents constantly throughout the day.
Antigravity vs. Gemini Code Assist vs. Jules
Google’s developer AI lineup isn’t just one product. If you’re a Google AI Pro or Ultra subscriber, you actually have three different coding tools available, and they’re built for different habits.
Frequently Asked Questions
Is Google Antigravity free?
Yes, there’s a $0 Individual plan with weekly rate limits, unlimited tab completions, and access to Google, Anthropic, and OpenAI models. Paid tiers ($20/month Pro, $100/month Ultra) raise the usage limits for heavier, all-day agent use.
What happened to Gemini CLI?
It was retired. Google announced the shift to Antigravity CLI on May 19, 2026, and Gemini CLI stopped working for Pro, Ultra, and free-tier users on June 18, 2026. Antigravity CLI is a Go-based successor that keeps most of Gemini CLI’s features as plugins.
Does Antigravity only work with Gemini models?
No. Alongside Gemini 3 Pro, Deep Think, and Flash, it also supports Claude Sonnet 4.6, Claude Opus 4.6, and GPT-OSS 120B, so you can pick a model per agent based on the task.
Is Antigravity the same as Gemini Code Assist?
No. Gemini Code Assist is an AI extension that plugs into an IDE you already use. Antigravity is a separate, standalone agent-first platform with its own editor, CLI, and SDK.
The Bottom Line
Antigravity is Google betting that the future of coding tools looks less like autocomplete and more like management: you describe the outcome, an agent plans and executes it, and you verify the result instead of writing every line yourself. The retirement of Gemini CLI in favor of Antigravity CLI wasn’t a rename, it was Google consolidating its entire developer AI strategy around this one platform.
If you’re still on Gemini CLI muscle memory, the free tier is the easiest way to see what changed. If you’re choosing between Google’s three coding tools, the honest answer is: Antigravity for anything you want handled end-to-end, Gemini Code Assist if you don’t want to leave your current IDE, and Jules for tasks you’re happy to queue up and forget about.
Sources: Google Developers Blog — Transitioning Gemini CLI to Antigravity CLI, Google Cloud — Gemini Code Assist overview, AI Wiki — Antigravity, Thinkpeak — Google Antigravity Pricing & Access 2026.