Google's agent-first development platform combining an AI IDE with multi-agent orchestration
Google Antigravity is an agentic development platform announced on the Google Developers Blog (November 2025). It pairs a familiar AI-assisted editor with a Manager Surface where developers spawn and observe agents working asynchronously across editor, terminal, and browser. Agents produce Artifacts—such as task lists, implementation plans, screenshots, and browser recordings—for review instead of relying only on raw tool logs. Public preview is offered at no cost for individuals on macOS, Windows, and Linux, with model choice including Gemini 3 Pro plus third-party models such as Claude Sonnet 4.5 and OpenAI GPT-OSS as described by Google.
Use cases
- Delegate end-to-end feature work where an agent edits code, runs the terminal, and verifies in the browser
- Run long-running bug reproduction and fix pipelines while monitoring progress asynchronously
- Iterate on UI changes with screenshot-backed updates for faster visual review
- Coordinate multiple agents on parallel maintenance tasks without blocking the main editor session
Key features
- Editor View: AI-powered IDE with tab completions and inline commands for hands-on coding
- Manager Surface: orchestrate multiple agents asynchronously across workspaces
- Cross-surface agents: planning and execution spanning editor, terminal, and browser
- Artifacts: review deliverables such as plans, screenshots, and recordings instead of only tool traces
- Knowledge retention: agents can save snippets and context for reuse on later tasks
- Model optionality: Gemini plus supported third-party models (per Google's announcement)
Who Is It For?
- Software engineers adopting agent-first workflows
- Teams evaluating Google's Gemini-backed tooling alongside other model providers
- Developers who want structured verification artifacts from autonomous agents
Frequently Asked Questions
- Where did Google announce Antigravity?
- Google introduced Antigravity on the Google Developers Blog under "Build with Google Antigravity, our new agentic development platform" (November 2025), with downloads referenced from antigravity.google.
- Is Antigravity only Gemini?
- Google describes model optionality with generous Gemini rate limits while also listing support for Anthropic Claude Sonnet 4.5 and OpenAI GPT-OSS in the launch post.
- What are Artifacts?
- Artifacts are tangible outputs agents attach for human review—such as plans, screenshots, or recordings—so you validate outcomes without parsing raw tool-call streams.
Related
Related
3 Indexed items
CodiumAI
CodiumAI is a platform focused on code analysis and testing, helping developers verify code correctness through automated test generation and analysis. The platform provides tools for understanding code behavior and generating relevant test cases. It integrates with popular IDEs to provide real-time feedback during development.
Cursor
Cursor is an AI-first code editor built on VS Code, launched in 2023 by Anysphere. It integrates AI capabilities directly into the coding workflow with features like code completion, natural language commands, and pair programming. Cursor 0.5 introduced Agent capabilities that can autonomously modify codebases. The editor supports Python, JavaScript, TypeScript, Go, Rust, and other major languages. It offers a free tier with 1000 code completions and paid plans for extended usage.
Replit Agent
Replit Agent is Replit's AI-powered autonomous coding agent launched in 2024, designed to build complete applications from natural language descriptions. Integrated directly into the Replit web-based IDE, the Agent can plan, write, test, and deploy applications independently. It leverages Replit's instant deployment capabilities to ship applications immediately after creation. The agent maintains project context and can take сотни lines of code actions autonomously.