⁠Google Antigravity

Google Antigravity introduces an Integrated Development Environment (IDE) designed for "agentic app building." This platform allows AI agents to autonomously plan, execute, and verify complex software tasks across various development interfaces, including code editors, terminals, and browsers. It’s built as a heavily modified fork of Visual Studio Code (VS Code), tailored for deep integration with Google’s proprietary tools and the Gemini AI model.

Google Antigravity: AI Agents for Software Development

Antigravity functions as an agentic development platform where AI agents can independently manage and complete complex software tasks. It features an Editor view, similar to a traditional IDE, and an Agent Manager view, which serves as a "mission control" for orchestrating multiple agents working in parallel. These agents generate "Artifacts" like task lists, implementation plans, screenshots, browser recordings, and code diffs. These artifacts act as verifiable deliverables, building user trust and facilitating feedback.

Enterprise Dev Teams and Cloud-Native Developers

This tool targets developers, particularly those within the Google Cloud ecosystem, aiming to enable anyone with an idea to bring it to fruition. Specific use cases include:

• Rapid Prototyping and MVP Development: Quickly build initial versions of applications.
• Offloading Maintenance: Automate long-running maintenance tasks.
• Safe Database Interaction: Query databases and explore schemas securely.
• Code Optimization: Improve code efficiency and performance.
• Security Vulnerability Identification: Pinpoint potential security flaws.
• Learning from Open-Source: Analyze and learn from existing open-source projects.

Gemini-Powered Agents with Cloud Integration

Antigravity is compatible with MacOS, Windows, and Linux. It supports Google’s Gemini 3 Pro, Gemini 3.5 Flash, and Gemini 3.1 models, alongside third-party options like Anthropic’s Claude Sonnet 4.5, Claude Opus 4.6, and OpenAI’s GPT-OSS. Key features include:

• Asynchronous Agents: Multiple agents can work in parallel.
• Browser Integration: Built-in capabilities for testing and verification within a browser environment.
• Self-Improvement: Agents learn from past work to enhance future performance.
• Feedback System: Users can comment directly on generated Artifacts.
• Command Line Interface (CLI) and Software Development Kit (SDK): For programmatic control.

The platform achieved 76.2% on the SWE-bench Verified benchmark, indicating its capability in automated software engineering tasks.

Google Cloud Pricing, Experimental Access

Google Antigravity is currently available for individuals at no charge during its public preview, offering "generous rate limits." For more intensive use, paid plans are available:

A credit system is in place for overages, with credits costing $0.01 each, though the specific credit-to-token conversion isn’t publicly disclosed. Pricing and usage limits have been subject to change since the tool’s launch.

Early-Stage, Limited Documentation, Google Cloud Lock-in

Despite its capabilities, Google Antigravity has several reported limitations. Users frequently complain about restrictive rate limits, particularly on the free tier, and a lack of transparency regarding usage calculation and credit conversion. There have been instances where initial generous limits were significantly reduced, leading to account lockouts even for paid subscribers. Performance issues are also common, including frequent "Agent terminated due to error" crashes, poor context retention over longer tasks, and noticeable lag, especially on macOS if GPU rasterization isn’t enabled. Furthermore, Google has reportedly suspended accounts of users who routed Antigravity’s backend through third-party tools, citing malicious usage without explicit terms of service violations. A notable concern for some users is the absence of an opt-out option for training on user data.