Google has repositioned Antigravity as a unified brand for AI coding harnesses fit for the agentic future. Originally introduced as a VS Code-style application during the dawn of the "vibecoding" era, Antigravity 2.0 puts agent orchestration front and center. Rather than performing one task at a time, developers can now set multiple agents to work on separate problems in parallel, multiplying engineering velocity.

The revamp brings deep integrations across the Google ecosystem. Developers can now seamlessly export prototypes from Google AI Studio directly into their local Antigravity desktop app, carrying over full context to continue development locally. Additionally, Google is launching a new Antigravity CLI to fully replace the Gemini CLI, allowing engineers to spin up and orchestrate agents straight from the terminal.

For custom setups, a new Antigravity SDK allows builders to create tailored agents optimized for Gemini models. To handle execution safely, Google is rolling out a "Managed Agents" feature within the Gemini API, providing agents with an isolated Linux environment to run tests and execute code securely. These agentic workflows are highly powered by the newly launched Gemini 3.5 Flash model, which handles complex coding benchmarks at a fraction of the cost and at 4x faster output speeds than previous frontier models.

Because running parallel coding agents is highly compute-intensive, Google has introduced a new $100/month tier for Google AI Ultra. This plan offers 5x higher usage limits in Antigravity compared to the standard AI Pro tier, ensuring developers don't hit immediate token bottlenecks. Furthermore, inspiration can now be captured on the move with the new Google AI Studio app for Android, letting developers prompt and prototype on mobile before exporting the code to Antigravity at home.