Google’s New Antigravity IDE: Revolutionize Your Coding for Free!

On Tuesday, November 18th, Google unveiled Gemini 3, its most advanced AI model, purportedly set to outperform OpenAI’s models in nearly all areas. Alongside this announcement, the company introduced updates to the CLI and a new development platform named Antigravity, meant to “evolve the IDE towards an agent-centric approach”.

Antigravity: Google’s New Agent-Based IDE

Google has made its intentions clear from the outset: it aims for Antigravity to “become the go-to platform for software development in the age of agents”. To achieve this, the company has designed a tool that is accessible to anyone who has an idea they want to bring to life. The tool maintains a familiar IDE experience, but enhanced with AI capabilities. Antigravity incorporates the models from Gemini 3, offering generous throughput limits for use on Gemini 3 Pro, as well as third-party models.

Google is setting high goals. The launch of Gemini 3 is expected to herald a new era in web development, where agents can operate truly autonomously for extended periods, utilizing multiple tools simultaneously.

We are not yet ready to operate continuously for days on end, but we are moving closer to a world where interaction with agents will occur at a higher level of abstraction, beyond mere requests and tool calls.

In Antigravity, Google introduces two complementary interfaces: an Editor that maintains the traditional IDE setup enhanced by AI, and a Manager designed to oversee multiple agents working in parallel. This increase in autonomy could significantly alter the developer’s role within the software production chain. As agents take on execution, testing, and some technical decision-making, the human role shifts towards setting goals, supervision, and arbitration.

The Four Core Principles of Antigravity

In its press release, Google outlines that Antigravity is built upon four complementary fundamentals.

1. Trust

Antigravity aims to establish a clear framework between the user and the AI agent. The tool groups the agent’s actions at the task level, allowing users to understand the reasoning behind its operations without displaying every technical detail. The tool generates tangible checkpoints like blueprints, screenshots, and recordings, giving users clear insights into how the agent is progressing and the basis of its decisions.

2. Autonomy

Google describes a new stage where the agent handles significant portions of technical work. The agent moves seamlessly between the editor, terminal, and browser as a human developer might, allowing it to execute multiple coherent actions without needing step-by-step approval. The introduction of the Manager space provides a way to oversee these activities.

3. Feedback

The interface is designed to facilitate ongoing dialogue between the human and the agent. Feedback can be given as text comments or annotations on visual elements, creating a more natural collaborative environment. These comments guide the agent throughout the process, allowing the user to adjust ongoing work without halting progress.

4. Self-improvement

Antigravity maintains a record of past actions to build an internal knowledge base. The agent uses this information to enhance consistency across projects. This mechanism improves the quality of responses and the stability of technical choices, mimicking a team that learns collectively from its experiences.

How to Access Antigravity

Google makes Antigravity available for free in a public preview, with compatibility across macOS, Windows, and Linux to reach the broadest audience.

The platform provides access to several AI models, including Gemini 3, as well as Claude Sonnet 4.5 and GPT-OSS.