Google has officially introduced Antigravity, a pioneering agentic development platform set to redefine how coding and software development are approached in the era of advanced AI models. Antigravity leverages the power of Google's most capable AI, Gemini 3, to offer developers an intelligent, autonomous environment for building software. This significant release positions Antigravity as the launchpad for a future where agents not only assist, but actively plan and execute complex programming tasks across multiple surfaces, bringing new dimensions of flexibility and automation to software creation.

The Vision Behind Antigravity

Google’s motivation for building Antigravity is clear: to make software development universally accessible and intuitive in an age dominated by agentic intelligence. Antigravity is designed as a central hub where anyone with an idea can transform it into reality. At present, the platform is available as a public preview, free for individuals, with generous access to Gemini 3 Pro and other top-tier models, including Anthropic’s Claude Sonnet 4.5 and OpenAI’s GPT-OSS.

A New Direction for IDEs: Agent-First Design

Antigravity advances beyond conventional Integrated Development Environments (IDEs), adopting an agent-first approach. This means that software development is now driven by autonomous agents capable of interacting with browsers, managing asynchronous tasks, and autonomously executing end-to-end workflows. The core experience remains familiar—an AI-powered editor—but Antigravity introduces innovative capabilities such as browser control, mission-like agent orchestration, and seamless multi-surface operation.

Antigravity’s Four Pillars

Trust: Antigravity tackles a key challenge in agent-powered coding: user trust. Unlike most products that either overwhelm users with every tool call or leave them in the dark, Antigravity contextualizes agent actions at a task level. Users can review artifacts—such as task lists, implementation plans, walkthroughs, screenshots, and browser recordings—making transparency and verification straightforward.

Autonomy: Agents in Antigravity can operate independently across different environments, from coding editors to browser windows and terminals. The platform introduces a unique 'Manager' surface, acting as a mission control center for spawning and observing agents in parallel workspaces—allowing developers to multitask and monitor progress asynchronously.

Feedback: Real-time feedback is woven throughout Antigravity. Users can comment and annotate across textual and visual artifacts (screenshots) without pausing agent operations. This ensures that agents are guided and improved, making the development process iterative and user-centric.

Self-Improvement: Antigravity incorporates continuous learning into its core. Agents draw from and contribute to an ever-growing knowledge base, aggregating useful code snippets, architectural patterns, and successful execution flows for future reference. This enables smarter agents over time and enhances productivity.

Experience the Future: Public Preview Now Available

With its agent-first architecture and tight integration of top AI models, Antigravity signals a new frontier in agent-assisted programming. Developers can now experiment freely with autonomous software development, enjoying immediate compatibility with MacOS, Linux, and Windows. The product is live for individuals to test—no charge, and a wide spectrum of model options ensures both flexibility and performance.