Google Antigravity, the new agentic development platform explained, what this agent platform does, what it costs and how to brief an AI agent

Developer working with Google Antigravity coding agents on a Microsoft Surface Laptop

Google Antigravity is Google’s agentic development platform where AI agents plan, write, run and test code while you supervise. It runs on Windows, macOS and Linux, according to Google’s official codelab (see further reading). Its agents work in an editor, a terminal (a command-line interface) and a browser, so you review results instead of typing every line. To build the skills behind this kind of work, explore the Certificate Program in Artificial Intelligence and Innovation from Founderz.

What you’ll take from this

You will know whether this tool fits your work and how to hand a task to an agent.

  • Tell the Agent Manager, the editor and the browser apart
  • Pick an AI model for each task
  • Check an agent’s work before you accept it
  • Write a brief the agent can verify

What is Google Antigravity and what is it used for?

Antigravity is a software development platform built around autonomous AI agents. You use it to delegate coding tasks, from writing code to testing it. You describe the job in plain language, and the agent works asynchronously while you do something else. It plans, writes code, runs commands in the terminal and checks the result in a built-in browser.

The shift is in your role. You stop typing every line and start directing agents. That is the skill Module 6 of the Certificate Program in Artificial Intelligence and Innovation from Founderz trains when it covers AI automations, workflows and agents. Google’s official codelab describes Antigravity 2.0 as the command center where you launch, supervise and organize your agents.

What does the agent-first era mean for coding with Antigravity?

It means the agent handles the multi-step work while you set goals and review results. Google presents Antigravity as a new agentic development platform for the agent-first era in its launch announcement (2025). A chat assistant hands back snippets. An AI agent in Antigravity edits files across the codebase, runs scripts and tests the outcome, planning and executing each step autonomously.

How does Antigravity work, and what is the Antigravity IDE?

The platform combines three connected surfaces: an agent control center, an editor and a browser the agent operates itself. Each covers one phase of the workflow, so you can delegate, review and test without leaving the app. The Antigravity IDE is the editor surface, an integrated development environment (IDE) derived from Visual Studio Code.

SurfaceWhat it doesUse it for
Agent ManagerControl center where you launch, orchestrate and supervise agents, including several in parallelLong tasks or several jobs at once
Editor (Antigravity IDE)IDE with code completion and inline commandsAdjusting code by hand
BrowserChrome controlled by the agent, with screenshots and browser recordingsTesting what the agent built

Which AI models does Google Antigravity use?

The platform lets you choose between Google and Anthropic models. According to Google Cloud (2026), Claude Opus 5.5 and Claude Sonnet 5.5 are available in Antigravity for Google Cloud customers, alongside Google Gemini models such as Gemini 3.8 Flash and Gemini 4 Argon. Availability depends on your plan, so check the model list inside the app.

These models are LLMs, large language models trained on huge volumes of text (see what an LLM is). Because the platform supports several, run the same coding task on two of them and compare the results. That shows which one handles your kind of work better.

Is Google Antigravity free, and can anyone use it?

Google has offered Antigravity free during its preview phase, with usage limits rather than a price tag. Today the available models and rate limits depend on your plan, and Google adjusts them often, so confirm current pricing on the official Antigravity site. For Google Cloud customers, partner models such as Claude are billed pay-as-you-go through your Google Cloud Billing account, according to Google Cloud (2026).

You sign in with a Google account. What you can use after that depends on the plan attached to it.

How do you install Antigravity?

You install it by downloading the desktop app for Windows, macOS or Linux and signing in with your Google account. Then you set the language in two places, the IDE interface and the agent’s replies.

  1. Download the installer from the official Antigravity site.
  2. Sign in with your Google account and choose a model.
  3. Open a project folder (a directory on your computer) as your workspace.
  4. Check settings and extensions for options such as language packs; exact paths vary by version.
  5. State the language and output format at the top of every brief.

What should you ask a Google Antigravity agent to do?

You can ask for a web page, a bug fix or the automation of a repetitive task, as long as the brief defines the result and how to check it. This example requests a simple page for a ceramics studio and forces the agent to show its plan before it starts coding.

Build a one-screen web page for a ceramics studio in Valencia, Spain. Include the studio name as the title, three courses with prices and a contact form. Before writing any code, show me the implementation plan and the task list. When you finish, open the page in the browser, take a screenshot and tell me what you checked. Reply in English.

The agent returns artifacts such as task lists, implementation plans, screenshots and browser recordings. If the plan looks wrong, comment on the artifact itself and the agent folds your feedback into its work, according to Google’s announcement (2025).

A common mistake is a brief like “make it look good” with no test attached. Name one check the agent can run, such as “the contact form shows a confirmation message after sending.”

Is Google Antigravity safe, and can you trust an agent’s artifacts?

Artifacts show what the agent did, but they do not prove it was done well, so safety depends on your review. Before you accept any job, check three things.

  • The code does what you asked, not only what the agent says it did.
  • The delivered result matches the initial plan.
  • The project holds no keys, passwords or personal data.

Frequently asked questions

Short answers to the questions beginners ask most.

Is Google Antigravity worth using if I don’t code?

It can be, with limits. Antigravity writes and tests code for you, but you still need enough understanding to review the result and spot errors. If you are starting from zero, practice on small projects and always ask for the plan before you accept any change.

What does the Antigravity 2.0 interface include?

It includes projects, conversations, skills, Model Context Protocol (MCP) servers and a customization section in settings, according to Google’s official codelab. Together these areas let you organize work by project and adjust how your agents behave during a task.

Does Google Antigravity have usage limits?

Yes. Rate limits and available models depend on your plan, and Google changes them frequently. Check the plan details on the official Antigravity site before a long project, because heavy agent work can reach a limit sooner than casual use.

Will Google Antigravity replace developers?

No. It delegates the execution, but someone still has to set the goal, review the code and take responsibility for the result. The developer moves toward directing and verifying agents. A non-technical professional can build prototypes with AI support, always with professional review before publishing.

What is Google’s AI service called?

Google’s AI models and assistant are called Gemini. Google Antigravity is a different product, a development platform that runs Gemini models and, depending on your plan, Anthropic’s Claude models. Gemini is the model family, and Antigravity is where agents use those models to write and test code.

Further reading

These sources cover Google’s official getting-started codelab and a reference entry.

To build the foundations behind agent work, see the Certificate Program in Artificial Intelligence and Innovation from Founderz, in collaboration with Microsoft. It is fully online and covers prompting, AI agents, automation with n8n and Copilot Studio, and responsible AI. It does not teach Antigravity itself. You receive the Microsoft + Founderz certification with official Microsoft badges. You can view the syllabus or request information.

Paul Delaney

Paul Delaney has been engineering AI prompts since the GPT-2 era, long before ChatGPT made prompting mainstream. Paul leads SEO, AEO, and GEO strategy at Founderz, improving how the school and its programs are discovered through traditional and AI-powered search. With more than 25 years of experience in education and digital growth, he has used AI daily since 2021 to support his commercial work.