Google unveiled Antigravity on November 17 alongside the Gemini 3 AI model as an AI-powered development platform designed to enable autonomous agents in software creation through an agent-first paradigm.
The platform functions as a fork of Visual Studio Code, allowing AI agents to operate across the code editor, terminal, and Chrome browser. This setup supports integration of multiple AI models, specifically Gemini 3 Pro, Anthropic’s Claude Sonnet 4.5, and OpenAI GPT-OSS. Google engineer Kevin Hou outlined the system’s core components as three main surfaces: an agent manager window for overseeing AI activities, a code editor for direct code manipulation, and an automated browser for web-based interactions.
Early user experiences highlight operational challenges during the free preview launch. Developers on Hacker News described rapid depletion of usage credits, with accounts exhausting limits after roughly 20 minutes of intensive tasks. One developer stated, “I start using it for my project and after about 20 mins—oh, no. Out of credits,” and subsequently returned to the competing platform Cursor due to the lack of options for purchasing additional quota.
Security concerns form a key aspect of the platform’s documentation. The terms of use state verbatim that “Antigravity is known to have certain security limitations,” with identified risks including data exfiltration and unauthorized code execution. In its standard setup, Antigravity runs in “agent-assisted development” mode, where the AI system independently determines which actions necessitate human approval before proceeding.
Account compatibility presents another hurdle for users. The platform encounters issues with Google Workspace accounts, prompting the FAQ to advise switching to personal Gmail addresses for access. Extension support draws from the Open VSX registry, which offers a narrower selection compared to the full Visual Studio Code marketplace available in other environments.
Google frames Antigravity as an initial preview of “agent-first software development,” positioning developers in the role of architects who orchestrate AI agents rather than engaging in direct code writing. The company plans to introduce Team and Enterprise pricing tiers in the future, while the ongoing viability of the current free tier faces questions related to the high computational demands of agentic workflows.
