Released by Google in November 2025, Antigravity is a development platform built around AI agents. It is based on a VS Code-style editor, so the interface feels familiar, but at its center sits an “Agent Manager” — a command center from which you can assign separate tasks to multiple AI agents at the same time. The agents do more than write code: they run terminal commands and drive a browser, and their work is presented in an organized form so you can trace what actually happened afterward. Through 2026, a CLI, an SDK, IDE extensions, and remote control were added, expanding it from a single desktop app into a platform with several entry points.
Key Features
- Parallel work through the Agent Manager: Run multiple local agents at once and manage them across workspaces. Conversations are grouped into Projects, so you can keep context separated per piece of work. Scheduled messages let you automate routine tasks on a recurring basis
- Moving between the editor and the agents: A full-featured VS Code-style editor lets you switch between writing code yourself and delegating to an agent within the same screen. It supports autonomous coding agents and shell command execution, and integrates with version control
- Agents that operate a browser: Browser-in-the-loop agents actually drive a browser to handle repetitive work. Checking whether the code you wrote behaves as intended on screen can be completed on the agent’s side
- Work is preserved as artifacts: An agent’s work is presented as organized artifacts. You can grasp what was done without scrolling endlessly back through a chat log
- A choice of models: In addition to the Gemini family, you can select Claude Sonnet / Opus, gpt-oss-120b, and others, matching the model to the nature of the task
- CLI, SDK, and extensions: There is an Antigravity CLI for terminal-centric work, an SDK for prototyping custom agents in Python, IDE extensions for use from your existing editor, and remote control for operating it while away
- MCP support: Through the Model Context Protocol you can connect external services and local environments, expanding the set of tools an agent can use
Pricing
| Plan | Price | What’s included |
|---|---|---|
| Individual | Free ($0) | Gemini models, Claude Sonnet / Opus, and gpt-oss-120b. Unlimited Tab completions and Command requests, with basic weekly rate limits |
| Google AI Pro | Follows the Google AI Pro subscription | Everything in Individual, plus more generous rate limits and a flexible AI credit pool |
| Google AI Ultra | Follows the Google AI Ultra subscription | Everything in Pro, plus even more generous rate limits and higher access to the latest Gemini models |
| Organization | Consumption-based (contact required) | Offered through Google Cloud. Google Cloud project integration and consumption-based pricing with the Gemini Enterprise Agent Platform. Included in select Gemini Enterprise subscriptions |
Rather than charging separately for Antigravity itself, rate limits are relaxed in line with Google’s AI subscriptions (Pro / Ultra). The official pricing page does not state specific amounts for Pro, Ultra, or Organization, so the actual cost has to be checked against the Google AI subscription pricing.
Pricing is current as of August 2026. Please check the official site for the latest information.
Pros & Cons
✅ Pros
- Even the free Individual plan lets you try the main features and models, so the barrier to entry is low
- Because it is based on a VS Code-style editor, you can move over without abandoning your existing habits and settings
- Running agents in parallel lets you advance tasks of different kinds — research, implementation, verification — at the same time
- Agent work is organized into artifacts, which makes reviewing delegated results easier
- Browser operation can be delegated too, so you avoid manually bouncing back and forth to check front-end behavior
- With a CLI, SDK, extensions, and remote control, there are multiple entry points to fit your existing workflow
⚠️ Cons
- The free plan has weekly rate limits, so serious daily use effectively assumes a Google AI subscription
- Parallel agents and the artifact model take some getting used to compared with a conventional IDE
- Because it autonomously edits code and runs shell commands, you need your own discipline around reviewing diffs and managing version control
- Organizational use goes through Google Cloud, so the contract and billing model differ from individual use
- It is a relatively new product with a fast pace of feature additions, so your working setup may shift over a short period
Comparison with Similar Services
| Criteria | Antigravity | Cursor | GitHub Copilot | Claude Code |
|---|---|---|---|---|
| Provider | Anysphere | GitHub (Microsoft) | Anthropic | |
| Form factor | Editor + Agent Manager + CLI | AI-first editor | Mainly an extension to existing editors | Terminal-centric agent |
| Parallel agents | Core feature | Supported | Supported | Supported |
| Browser control | Supported out of the box | Via extensions / integrations | Limited | Via integrations |
| Free plan | Yes | Yes | Yes | Assumes a paid plan |
| Model choice | Gemini / Claude / gpt-oss | Multiple models | Multiple models | Claude family |
Who Is It For
- Developers who want to hand multiple tasks to AI at once and concentrate on review and judgment themselves
- People who want to try an agent-first workflow while keeping the feel of a VS Code-style editor
- Those who want to handle front-end implementation and on-screen verification together instead of switching back and forth
- Anyone already subscribed to Google AI Pro / Ultra who wants coding support within that same allowance
- Engineers who want to assemble their own agents from the CLI or SDK
- People who want to see what agent-first development is like, for free, before committing
Summary
Antigravity is designed less as “an editor that completes your code” and more as “a command center that distributes work across multiple agents.” Being based on a VS Code-style editor makes the move over straightforward, and the free Individual plan — which includes Gemini and Claude models — is a friendly entry point. That said, since the tool autonomously rewrites code and executes commands, you need to keep your own discipline around diff review and version control. A realistic path is to start on the free plan with smaller tasks, then consider a Google AI subscription once the rate limits start to feel tight.