Antigravity turns its agents’ responses into interactive interfaces

Google Antigravity can now generate interactive HTML interfaces, charts, and simulations directly within its conversations.

Google is expanding Antigravity with a system of interactive artifacts designed to move beyond the limitations of text responses, static images, and Mermaid diagrams. Agents within the company’s development environment can now produce visual components, dynamic dashboards, interactive charts, and explanatory simulations, then display them directly inside the conversation.

Introduced with Antigravity version 2.11.0, the feature relies on content generated with HTML, CSS, and JavaScript. Users can interact with the result directly in the conversation or open it in the dedicated artifacts panel. The idea is to make concepts that would be difficult to communicate through a written description immediately visible and testable.

A developer could ask the agent to map a project’s dependencies, create a dashboard from a dataset, simulate the behavior of a software architecture, or prototype an interface component. Google’s announcement also mentions interactive 3D simulations, although the company has not yet specified which technologies or formats are supported for this particular use case.

The workflow remains integrated into the conversation. Users can enter a natural-language request or run the `/generativeui` command. The agent then builds the artifact, displays it in a preview area, and can revise it in response to additional instructions. Users can inspect individual elements, request changes, and export the result as a standalone HTML file.

Antigravity officially supports mathematical notation through KaTeX, along with charts built using Chart.js and Plotly. This makes it possible to generate time-series graphs, comparisons, statistical visualizations, and dashboards without having to configure those tools manually beforehand.

Google also presents these visualizations as requiring no additional dependency installation. They are rendered locally and can continue to work offline once exported. However, this does not necessarily mean that the generated code uses no libraries. Antigravity provides built-in support for several components, primarily removing the need for users to add them to their own projects.

The feature gradually turns an agent response into a temporary mini application. Instead of receiving only an explanation of a codebase or the evolution of a dataset, developers can explore the result, change parameters, and observe the consequences. This approach could reduce the need to switch constantly between a terminal, a visualization platform, and an interface editor.

These artifacts should not be considered production-ready components without further review. As with any generated code, their contents need to be checked before being integrated into a product or shared publicly. Accessibility, browser compatibility, performance, data accuracy, and the handling of sensitive information remain the developer’s responsibility.

Google’s announcement also provides no detailed information about network restrictions or the isolation model applied to JavaScript running inside the preview. HTML export makes the artifact easier to reuse, but developers still need to review external resources, links, and any data embedded in the file.

Interactive Generative UI artifacts are available in Antigravity 2.11.0. Google notes that new releases are deployed gradually and may take several days to reach every user. The feature reflects a broader shift among development assistants: their responses are no longer intended only to be read, but also to be explored, tested, and reused.