Google Antigravity: new AI agent-assisted software development platform

The company has announced Google Antigravity, a platform that redefines software development through artificial intelligence agents. The system is based on Gemini 3 Pro and is available in public preview at no charge, with generous model usage limits.

Antigravity integrates two distinct work environments. The first is an editing interface that functions as a traditional development environment, with autocompletion, inline commands, and a functional agent in the side panel. The second component is a management surface designed for asynchronous interaction, allowing users to launch and monitor multiple agents working in parallel across different workspaces. This architecture responds to current models' capability to operate for extended periods without human intervention.

The system organizes agent actions into tasks with high-level progress summaries. During execution, agents generate artifacts such as task lists, implementation plans, screenshots, and browser recordings, elements designed to facilitate validation of completed work. The platform incorporates verification mechanisms where the agent evaluates its own work before presenting it.

Antigravity allows providing feedback on any surface and artifact while the agent continues operating. Users can add comments on text documents or mark specific sections in screenshots, information that automatically integrates into the execution process without stopping the agent.

The platform includes a knowledge management system where agent actions contribute to a database. This repository stores code snippets, derived architectures, and step sequences to complete specific subtasks, information that the agent can retrieve in subsequent work.

Google Antigravity is compatible with MacOS, Linux, and Windows. The system offers access to Google's Gemini 3, Anthropic's Claude Sonnet 4.5, and OpenAI's GPT-OSS within the agent, allowing developers to select among different models according to their needs.

Key points

• - Google introduces Antigravity, AI agent-based software development platform with Gemini 3 Pro
• - Integrates two environments: traditional editing interface and asynchronous management surface for multiple agents
• - Agents generate artifacts such as implementation plans, screenshots, and browser recordings
• - Allows providing real-time feedback without stopping agent execution
• - Includes knowledge management system that stores code and architectures for future work
• - Available in public preview at no charge with generous usage limits
• - Compatible with MacOS, Linux, and Windows
• - Offers access to Gemini 3, Claude Sonnet 4.5, and GPT-OSS within the agent