Google Anti-Gravity — AI Coding Tool

Google Anti-Gravity — AI Coding Tool by Google

Google Anti-Gravity (officially styled “Antigravity”) is Google’s agentic coding platform, launched in public preview on November 18, 2025 alongside Gemini 3 Pro. Built as a VS Code fork with an “agent manager” surface, it brings Google’s most powerful AI capabilities directly into the development workflow. Anti-Gravity 2.0, released May 2026, switched the default model to Gemini 3.5 Flash — Google’s strongest agentic/coding model at the time, outperforming Gemini 3.1 Pro on coding and agentic benchmarks.

Antigravity competes directly with tools like Claude Code, OpenAI Codex, Cursor, and AWS Kiro. Google’s model lineup moves fast — check the Google Models guide for the current Gemini generation, since Antigravity’s default model gets updated as new Gemini releases ship.

What Makes Antigravity Different

Google’s Antigravity leverages advantages unique to Google’s AI stack:

Gemini 3.x backbone — very large context window (historically among the largest of any coding AI)
Google Search grounding — can look up live documentation, Stack Overflow, GitHub issues
Google Cloud integration — native awareness of GCP services, BigQuery, Cloud Run, GKE
Entire repo understanding — can load and reason over very large monorepos
Multimodal input — understands diagrams, screenshots, error screenshots as input

Antigravity 2.0 Key Features

Massive Context

With Gemini’s large context window, Antigravity can read:

An entire large monorepo in one pass
Long migration histories
Multiple related repositories simultaneously
All tests, configs, documentation, and source code at once

Google Search Grounding

When Anti-Gravity encounters an unfamiliar library, deprecated API, or error message, it can search Google in real time and incorporate up-to-date answers — something no purely local model can do.

Agentic Code Edits

Like Cursor, Windsurf, and Claude Code, Anti-Gravity can:

Make changes across multiple files
Run shell commands and test suites
Open pull requests on GitHub
Iterate on failing tests automatically

GCP-Native Features

For teams building on Google Cloud:

Understands GCP architecture patterns
Generates Cloud Run, GKE, and Cloud Functions configs
Can query BigQuery schema and generate SQL
Integrates with Google Cloud Build for CI/CD

Anti-Gravity vs Competing Tools

Feature	Anti-Gravity	Claude Code	Codex	Kiro
Context window	Large (Gemini 3.x)	200K (Claude)	Large (GPT-5.x)	~200K
Search grounding	Yes (Google)	Yes (Web)	No	No
Cloud integration	GCP-native	AWS via CLI	Via GitHub	AWS-native
Agentic editing	Yes	Yes	Yes (cloud sandbox)	Yes (spec-driven)
Multimodal input	Yes	No	No	No
Available in	Browser + CLI	Terminal	Cloud + CLI	IDE

Google Cloud Integration

Anti-Gravity understands Google Cloud services natively:

GCP Service	What Anti-Gravity Can Do
Cloud Run	Generate Dockerfiles and service configs
GKE (Kubernetes)	Write Kubernetes manifests
BigQuery	Generate SQL from natural language
Cloud Functions	Write and deploy serverless functions
Firestore	Model data structure and generate queries
Cloud Build	Generate CI/CD pipeline configs

Getting Started

Anti-Gravity 2.0 is available via:

Google AI Studio (aistudio.google.com) — Browser-based interface
Gemini API — For developers integrating into workflows
Google Cloud Console — Integrated into GCP developer tools
IDE Extension — For VS Code and JetBrains (via Google Cloud Code extension)

Anti-Gravity and Gemini CLI

Anti-Gravity’s capabilities are also accessible via the Gemini CLI for terminal users:

# Install Gemini CLI
npm install -g @google/gemini-cli

# Authenticate
gemini auth login

# Use with your codebase
cd my-project
gemini "Refactor the database module to use connection pooling and add tests"