Google Antigravity is an AI-powered agentic software development platform. Rather than functioning as a standard code completion tool (like GitHub Copilot or traditional IDE extensions), Antigravity shifts software creation to an "agent-first" paradigm where autonomous AI agents handle end-to-end task execution.
Antigravity Manager: A "mission control" interface used to spawn, orchestrate, and observe multiple AI agents running concurrently across background tasks.
Antigravity IDE: A heavily modified, AI-first editor environment built on Visual Studio Code architecture, offering side-by-side execution, full codebase context, and agent management.
CLI & SDK Ecosystem: Terminal interfaces and developer SDKs for programmatically triggering, automating, and integrating Antigravity agents into larger workflows.
IDE Extensions: official integrations allowing Antigravity's autonomous agent engine to run within third-party environments such as VS Code, JetBrains IDEs, and Zed.
Autonomous Task Execution : Developers specify goals in natural language. Agents plan steps, generate or refactor code, run terminal commands, and launch browser sessions to test solutions independently.
Integrated Tool Access: Agents hold direct access to the local machine’s editor, terminal, and browser capabilities to install dependencies, run tests, and debug UI elements.
Verifiable Artifacts: To maintain transparency, agents generate intermediate "Artifacts"—including task plans, implementation summaries, screenshots, and browser session recordings—before merging changes.
Multi-Model Support: Powered primarily by Google's Gemini models (Gemini 3 / 3.1 series), with support for external partner and open models.