Google Antigravity

What is Google Antigravity?

Google’s documentation describes Antigravity as an agentic, Gemini-powered browser IDE. It runs entirely in the browser and provides a code editor with integrated AI assistance. The distinguishing feature is task-budgeted autonomous coding flows: a developer can give the tool a goal (e.g., “refactor this module” or “fix these three bugs”) and Antigravity will independently plan and execute a series of edits, constrained by a configurable budget of steps or time. The underlying agent uses Google’s Gemini model to break down the request and implement the changes. Unlike a simple autocomplete or chat sidebar, Antigravity operates in a more autonomous mode, showing each proposed change as it works. The repository and public documentation indicate that the IDE supports common language runtimes, Git integration, and terminal access within the browser, though the agentic flows are the primary differentiator.

Who is Google Antigravity for?

Antigravity appears aimed at developers who are willing to delegate repetitive or well-defined coding tasks to an AI agent. The docs present use cases like boilerplate generation, bug fixes with clear reproduction steps, and small-scale refactoring. It is not positioned as a replacement for human judgment on complex architecture decisions. The browser-based nature means it works on a range of devices without local setup. The agentic flows may particularly suit engineers who need to move fast on straightforward changes and are comfortable reviewing AI-generated code before committing. Given the task-budget model, developers have a measure of control over how much autonomy the tool gets before asking for confirmation or stopping. It is likely less suited for teams that require full off-line operation or low-level system programming outside the browser environment.

How much does Google Antigravity cost?

The pricing page lists a freemium model. A free tier is available, but details on its limits (such as number of agentic runs per day or Gemini model access) are not specified in the public materials reviewed. The paid plan starts at $20 per month. The pricing page suggests that this subscription unlocks higher budgets for autonomous flows, access to more capable Gemini models, and possibly expanded storage or collaboration features. No annual billing options or enterprise pricing are mentioned on the public site. The $20/mo price point is consistent with other premium AI coding assistants on the market.