Alongside today’s announcement of Gemini 3 Pro, Google has revealed Antigravity, a development tool that uses Gemini 3 Pro, along with other third-party models. Google says that Antigravity, which supports multiple agents and gives them direct access to the editor, terminal, and browser, is designed for an “agent-first future.”
One of the key components of Antigravity is how it reports on its own work. As it completes tasks, it will produce what Google calls Artifacts: task lists, plans, screenshots, and browser recordings that are intended to verify both the work it’s done and what it will do. Antigravity will also report on its actions and external tool use along the way, but Google says that Artifacts are “easier for users to verify” than full lists of a models’ actions and tool calls.
Antigravity’s other big change is that it offers two main usage views. The default Editor view offers a familiar Integrated Development Environment (IDE) experience, similar to rivals like Cursor and GitHub Copilot, with an agent in a side panel. The new Manager view is instead designed for controlling multiple agents at once, allowing each to work more autonomously. Google compares it to “mission control for spawning, orchestrating, and observing multiple agents across multiple workspaces in parallel.”
Google has introduced more ways to give feedback to AI agents as they work, with the ability to leave comments on specific Artifacts for an agent to take into account without breaking up its work to do so. The company also says that agents in Antigravity will be able to “learn from past work,” retaining specific snippets of code or the steps required to carry out certain tasks.
In this demo, Antigravity builds a basic flight tracker app, tests it, and reports on that test with a browser recording.
Antigravity is available in a public preview now, compatible with Windows, macOS, and Linux. It’s free to use, with what Google calls “generous rate limits” for Gemini 3 Pro, though it also supports Claude Sonnet 4.5 and OpenAI’s GPT-OSS. Google says rate limits refresh every five hours, and that only “a very small fraction of power users” will ever hit the limits.
