Overview
- Antigravity lets autonomous agents plan, write, run, test, and verify software by operating across the editor, terminal, and an integrated browser.
- The platform ships with two surfaces: a familiar Editor view and a Manager view that functions as mission control for spawning and coordinating multiple agents.
- Google supports model choice inside Antigravity, including Gemini 3 Pro, Anthropic’s Claude Sonnet 4.5, and OpenAI’s GPT‑OSS, subject to capacity.
- The system emphasizes trust through Artifacts such as task lists, implementation plans, screenshots, and browser recordings, with asynchronous comments and a knowledge base that lets agents learn from past work.
- The public preview is free on Windows, macOS, and Linux with rate limits that refresh every five hours; early users report credits depleting quickly and Dataconomy notes terms warning of known security limitations, while some outlets report VS Code‑based compatibility and Open VSX extensions.