Google Antigravity makes Ide A Story Plane for agentitic coding

Google launched antigravity as an agentic development platform on top of gemini 3 Antigravity launched on November 18, 2025, alongside Gemini 3 as part of Google's push for Agent Centric developer tools.

What's the real deal?

Antigravity is described by Google as a new platform for agentic development at its core. The goal is to appear the agent in the first future of the agent, with browser control and asynchronous communication patterns and patterns that allow agents to organize and complete the end of software tasks.

In practice, existing intelligence looks and behaves like a modern AI programmer but treats agents like first-class employees. Agents can execute tasks, communicate with other agents, edit files, run commands, and call a browser. The engineer works at the task level, while the system handles the interaction of the lower tools.

Under the hood, antigravity is an electronic application based on Visual Studio code. It requires a Google Account login and ships as a free public preview for macos, Linux, and windows.

Models, pricing, and runtime environment

Antigravity presents multiple base models within the framework of the same agent. In current testing, agents can use Gemini 3, Anthropic Claude Sonnet 4.5, and Opelai GPT OSS models. This gives developers the option to model within one long ede instead of tying them to a single vendor.

For some users, the antigravity feature is available free of charge. Google describes the use of the Gemini 3 Pro as under the generous limits of the rate that refreshes every 5 hours, and it is expected that there will be a small fraction of power users to hit it.

Editor's View and Administrator's View

Antigravity presents 2 main modes of activity corresponding to different neural models. Texts and coverage often describe these as songs from an editor's view and a manager's view.

Editor view is the default. It looks like a standard endo with an agent on the side panel. The agent can read and edit files, suggest changes inline, and use a terminal and browser when needed.

The Manager view suggests issuing from single files to multiple agents and workstations. This is where you program a few agents rather than line-by-line code.

Art, not lumber for raw tools

The main object of antigravity is the artifact system. Instead of only exposing raw logs with tools, agents produce human-readable art that summarizes what they do and why.

Artifacts have organized items that can include task lists, implementation strategies, Walkaugh documents, screenshots, and browser recordings. They represent functionality at the function level rather than at the API driver level and are designed to be easier for developers to verify than to follow model traces.

Google positions this as an answer to the trust problem in current agent frameworks. Many tools show all the internal steps, which affect users, or hide them all and show only the final unique code. Antigravity tries to stay in the middle through the use of surfacing Task Ariffacts and sufficient authentication signals for the engineer to study what the agent has done.

Four Channels of Design and Feedback

Antigravity is clearly built around 4 tenots, trust, independence, feedback, and self-improvement.

Trust is handled through creative measures and affirmative measures. Autonomy comes from giving agents access to multiple locations, the editor, the end of the browser, so they can run complex workflows without constant prompting. Feedback is enabled by commenting on artifacts, and self-improvement is combined with agents learning from past work and reusing successful practices.

Antigravity allows developers to annotate directly on specific objects, including text and screens. Agents can incorporate this feedback into their ongoing work without abandoning the current run. This allows you to correct partial inconsistencies without restarting the entire operation.

The platform also exposes a knowledge feature where agents can save code snippets or sequences of steps from previous tasks. Over time, this becomes an internal playbook that agents can ask, rather than re-discovering the same strategies for each new project.

Key acquisition

1. Antigravity is the first development platform that turns ide into a control plane where agents work across the editor, browser environment and browser, instead of a small inline assistant.
2. The program is a virtual fork of Studio Code that runs as a free public preview for Windows, Macos and Linux, with the Gemini 3 Pro rating restrictions and the optional Claude Sonnet 4.5 and GPT OSS.
3. Antigravity features 2 main modes, the editor view for manual coding with the Sidebar agent bar and the Manager View as a deployment control connector to deploy asynclestrates.
4. Agents emit artifacts, task lists, execution plans, screenshots, screen recordings, browser recordings and more, making it a proven proof of performance for raw logs and enabling asynchronous workflows.
5. Feedback and improvements are done internally, developers can attach Google Docs-style comments to Artifacts everywhere, and agents enter this feedback and learn from the development knowledge base without starting jobs.

Google Antigravity is a pragmatic step in agentic development. Anchors Gemini 3 Pro within the original workflow of Ide, presents the editor's view and the manager's view of the agent's managers, and includes the ability to visualize tasks through creativity. The four sets, trust, independence, feedback, self-improvement, are removed from proven fields and persistent information rather than opaque. Overall, antigravity treats the endone as a sensitive area for independent agents, not a dialog window with coded actions.

Look Full technical details here. Feel free to take a look at ours GitHub page for tutorials, code and notebooks. Also, feel free to follow us Kind of stubborn and don't forget to join ours 100K + ML Subreddit and sign up Our newsletter. Wait! Do you telegraph? Now you can join us by telegraph.