Overview
Google Antigravity is an AI-powered integrated development environment introduced by Google on November 18, 2025, alongside its advanced AI model Gemini 3. Built as a fork of Visual Studio Code, Antigravity represents a fundamental shift in software development by enabling developers to delegate complex coding tasks to autonomous AI agents that can plan, execute, and validate entire development workflows.
The platform is designed for developers who want to accelerate their coding productivity through intelligent automation. Rather than providing simple code suggestions, Antigravity allows AI agents to autonomously handle multi-step tasks across the editor, terminal, and integrated browser while maintaining full transparency through generated artifacts.
Antigravity is currently available in public preview at no cost for individual developers, with desktop builds for macOS, Windows, and supported 64-bit Linux distributions (requiring glibc ≥ 2.28 and glibcxx ≥ 3.4.25). Preview access commonly starts with personal Google accounts, but Google has also referenced Workspace-linked tiers with different quotas and eligibility that may vary by geography and admin settings. The platform supports multiple AI models including Gemini 3 Pro, Anthropic's Claude Sonnet 4.5, and OpenAI's GPT-OSS open-weight models, giving developers flexibility in choosing the right model for each task.
Key Features
Agent-First Architecture — Shifts from traditional AI code completion to autonomous agents that plan, execute, and validate entire development tasks, allowing developers to focus on higher-level design and decision-making.
Dual Interface System — Combines a familiar VS Code-style Editor View for coding with a Manager view for orchestrating multiple agents working simultaneously across different projects and workspaces, with progress tracked through artifacts.
Smart Artifacts Generation — Automatically creates comprehensive Markdown artifacts including task lists, implementation plans, code diffs, screenshots, and browser recordings, providing transparent insights into agent activities. Actual execution and approval behavior depends on configured policies (e.g., review-driven vs agent-driven modes, terminal approval settings, secure mode).
Cross-Surface Integration — Provides agents with direct access to the code editor, terminal, and browser integration via a browser tool/subagent for end-to-end testing and web interactions, governed by configurable policies (e.g., JavaScript execution and review/approval rules).
Multi-Agent Orchestration — Manages multiple agents working asynchronously across different projects, enabling efficient handling of large codebases and parallel task execution.
Multi-Model Support — Supports Google's Gemini 3 Pro alongside Anthropic's Claude Sonnet 4.5 and OpenAI's GPT-OSS open-weight models, allowing developers to select the optimal model for specific tasks.
Pricing & Plans
Google Antigravity is currently in public preview with an Individual plan at $0/month. This free access includes:
- Multi-agent support within generous per-window rate limits
- Integrated Gemini 3 Pro access in the IDE
- Core features including tab completions and key surfaces (Agent Manager, Browser integration)
- Complete integration with all development surfaces (editor, terminal, browser)
- Support for Windows, macOS, and Linux distributions (requires glibc ≥ 2.28 and glibcxx ≥ 3.4.25; e.g., Ubuntu 20+, Debian 10+, Fedora 36+, RHEL 8+)
Rate Limits: During preview, free users are subject to a weekly usage quota, while Google AI Pro/Ultra subscribers get priority access with quotas refreshing about every five hours. Limits are based on "work done" by agents rather than raw request counts. Google expects most individual developers not to hit these quotas in typical workflows, but heavy multi-agent runs can reach the caps.
Future Pricing: Google now references additional tiers tied to Google One / Google Workspace / Google Cloud (such as Workspace-linked plans with different quotas), but exact feature entitlements and final packaging may still change during preview.
Note: Gemini API pricing applies to API usage (e.g., via AI Studio/Vertex AI), not necessarily to Antigravity's in-IDE quotas. Refer to Google's official Gemini pricing for up-to-date token rates and context-window tiers.
Pros & Cons
Pros:
- Free Individual plan ($0/month) during public preview with core functionality
- Agent-first design enables autonomous handling of complex, multi-step coding tasks
- Transparent artifact system builds trust by showing agent reasoning and planned changes, with configurable review policies
- Built on a VS Code-derived foundation, offering a familiar interface and supporting many existing VS Code extensions, though some plugins may require manual installation via VSIX or are not yet fully compatible
- Multi-model support provides flexibility in choosing AI capabilities
- Cross-platform support for Windows, macOS, and Linux distributions (glibc ≥ 2.28, glibcxx ≥ 3.4.25)
Cons:
- Still in public preview with known security limitations and potential stability issues
- Account eligibility can be complex; availability may vary by geography and admin settings
- Weekly usage quotas for free users; heavy multi-agent usage can exhaust limits until reset
- Newer platform with smaller community compared to established IDEs
- Requires learning new agent orchestration paradigm
- Uses OpenVSX/VSIX ecosystem instead of Microsoft's extension marketplace; users should verify publishers for recommended extensions due to potential supply-chain risks
Best For
- Developers working on complex projects requiring multi-step automated workflows
- Teams looking to accelerate development velocity through autonomous AI assistance
- Engineers comfortable with agent-based development and willing to adopt new paradigms
- Projects that benefit from parallel task execution across multiple codebases
- Developers seeking transparent AI assistance with verifiable artifacts and decision trails
- Early adopters willing to explore cutting-edge AI development tools during public preview
FAQ
Who can use Antigravity during the preview?
Antigravity is currently available as a desktop IDE for macOS, Windows, and selected 64-bit Linux distributions (requiring glibc ≥ 2.28 and glibcxx ≥ 3.4.25, such as Ubuntu 20+, Debian 10+, Fedora 36+, RHEL 8+). Preview access commonly starts with personal Google accounts, but Google has also referenced Workspace-linked tiers (e.g., AI Ultra for Business) with different quotas and eligibility. Availability may vary by geography and admin settings.
Is Google Antigravity free to use?
Yes, Antigravity currently offers an Individual plan at $0/month during its public preview phase. Core features remain available across tiers (including tab completions and key surfaces like Agent Manager and Browser integration), with weekly usage quotas for free users. Google AI Pro/Ultra subscribers get priority access with higher quotas that refresh about every five hours. Quotas are based on "work done" by agents rather than raw request counts. Google has referenced additional tiers tied to Google One / Google Workspace / Google Cloud, but exact feature entitlements and final pricing may still change during preview.
What operating systems does Antigravity support?
Antigravity is available for Windows, macOS, and Linux platforms. On Linux, it requires glibc ≥ 2.28 and glibcxx ≥ 3.4.25 (e.g., Ubuntu 20+, Debian 10+, Fedora 36+, RHEL 8+). The IDE is built on a VS Code-derived foundation, ensuring broad compatibility across operating systems.
Which AI models does Antigravity support?
Antigravity primarily uses Google's Gemini 3 Pro model but also supports Anthropic's Claude Sonnet 4.5 and OpenAI's GPT-OSS open-weight models. This multi-model approach allows developers to choose the most suitable model for specific tasks.
How does Antigravity differ from traditional AI coding assistants?
Unlike traditional AI tools that provide code suggestions or completions, Antigravity uses autonomous agents that can plan, execute, and validate entire development tasks. Agents have access to the editor, terminal, and browser integration, enabling them to handle complex multi-step workflows independently. Artifacts provide verifiable deliverables (plans, diffs, screenshots, recordings) for review, with actual execution and approval behavior depending on configured policies (e.g., review-driven vs agent-driven modes, terminal approval settings, secure mode).
Are there rate limits during the public preview?
Yes. During preview, free users are subject to a weekly usage quota, while Google AI Pro/Ultra subscribers get priority access with quotas refreshing about every five hours. Limits are based on "work done" by agents rather than raw request counts or lines of code. Under normal individual workloads you're unlikely to hit them, but intensive multi-agent tasks can exhaust the quota until it resets.
Is my code and data secure with Antigravity?
Treat Antigravity as preview software: enable Secure mode and stricter review policies, require explicit approval for terminal commands, and limit browser automation on untrusted inputs. Security researchers have demonstrated indirect prompt injection leading to data exfiltration via browser subagents, so avoid highly sensitive repositories until controls and guidance mature. You can configure security settings including Review policy, Terminal execution policy, and JavaScript execution policy. Review the Antigravity Terms and Generative AI Terms of Service, and carefully configure data-sharing settings before use.
Can Antigravity agents work on multiple projects simultaneously?
Yes, the Manager view allows developers to orchestrate and monitor multiple agents working across different projects and workspaces simultaneously, with progress tracked through artifacts. This enables asynchronous task execution and efficient handling of large codebases.
Does Antigravity work with existing VS Code extensions?
Because VS Code forks may not use Microsoft's extension marketplace, Antigravity relies on VS Code-compatible ecosystems like OpenVSX/VSIX. Many popular VS Code extensions work out of the box or via manual VSIX installation, but not every Marketplace extension is visible or fully compatible. Users should verify publishers for recommended extensions—recent research showed "missing extension" recommendations could create supply-chain risk, which vendors have since worked to fix. You should test mission-critical plugins during the public preview to ensure they work as expected.
