News
Feb 19, 2026
Tech Updates
Enterprise
Artificial Intelligence
Americas
Google Launches Antigravity: AI Agents Take Control of Development
NewDecoded
3 min read
Image by Google
Google released Antigravity on November 18, 2025, marking its entry into the agentic development platform space. Built on Gemini 3, Google's most advanced coding model, the platform enables AI agents to autonomously handle complex software tasks from planning through testing. The free public preview supports MacOS, Linux, and Windows with generous rate limits that refresh every five hours.
Unlike traditional coding assistants that offer autocomplete suggestions, Antigravity fundamentally reimagines the IDE around autonomous agents. The platform provides two complementary interfaces: a familiar Editor view with tab completions and inline commands, and an innovative Agent Manager that functions as mission control for orchestrating multiple agents across workspaces simultaneously. Agents can write code, launch terminals, actuate browsers, and verify their own work without constant human intervention.
The platform operates on four core principles that address current limitations in AI-assisted development. Trust comes through task-level abstractions and artifacts like implementation plans, walkthroughs, and browser recordings that help users validate agent work. Autonomy allows agents to operate across multiple surfaces at once, while feedback mechanisms let developers comment directly on artifacts without stopping execution. A knowledge base system enables self-improvement, letting agents learn from past work and successful patterns.
Developers can choose between three operational modes depending on their comfort level: fully autonomous "Autopilot" where agents work independently, review-driven mode requiring permission for most actions, or the recommended agent-assisted approach that balances control with automation. The platform includes access to multiple AI models beyond Gemini 3, including Anthropic's Claude Sonnet 4.5 and OpenAI's GPT-OSS, giving developers model flexibility for different tasks.
Google envisions Antigravity as "the home base for software development in the era of agents," with the ultimate goal of enabling anyone with an idea to build it into reality. The company plans frequent feature releases and has established documentation and use case libraries to help developers explore the platform's capabilities.
Google's Antigravity launch represents a strategic response to the competitive AI coding assistant market, where GitHub Copilot dominates and newer entrants like Cursor and Replit's Ghostwriter have gained traction. By positioning Antigravity as an agent-first platform rather than a code completion tool, Google is betting on a future where developers orchestrate AI agents asynchronously rather than coding line-by-line.
The dual-interface approach (Editor and Agent Manager) signals Google's belief that current synchronous IDE paradigms won't scale as models become capable of longer autonomous work sessions. The free tier with generous rate limits mirrors the land-grab strategy seen across AI tools, though the intelligent rate limiting based on task complexity rather than simple prompt counts suggests Google is managing compute costs while encouraging exploration.
Most significantly, the multi-model support (including competitors Claude and GPT-OSS) positions Antigravity as a platform play rather than purely a Gemini showcase, potentially creating ecosystem lock-in even as users switch between underlying models.
