I currently have an individual Google AI subscription and use Antigravity. My usual account access is from the United States, but I am planning to set up a private development automation workflow on an Ubuntu EC2 instance in AWS Sydney, Australia.
Before proceeding, I would like to clarify the permitted use of Antigravity and understand the recommended alternative for unattended coding automation.
My intended workflow is:
A private GitHub repository contains the project.
A server-side scheduler selects explicitly approved GitHub Issues.
A coding agent edits code in an isolated working directory, runs tests, and prepares a draft pull request for human review.
The workflow is private, for my own projects, and is not exposed as a public or multi-user service.
Could you please clarify the following?
Is it permitted to run the official, unmodified Antigravity CLI on a remote AWS EC2 server using my individual subscription, where the server is in Australia but my usual account access is from the United States?
Does the answer change if the CLI is invoked non-interactively by a private scheduler, rather than used manually in a terminal?
Are there any restrictions concerning the server’s AWS region, account location, or changes between supported AWS regions such as Australia, Japan, and South Korea?
I understand that the Antigravity Terms restrict third-party tools and services from accessing Antigravity. Is there an officially supported way to orchestrate the unmodified CLI, or should third-party coding agents use a separately configured Gemini API key instead?
If I use OpenCode with a Google AI Studio Gemini API key, is that the recommended supported route for this type of private automation?
I would like to use a supported configuration and avoid unintentionally violating the applicable terms. Thank you for clarifying the applicable rules and recommended setup.
Welcome to the Forum, and thank you for reaching out to check on this before setting up your workflow.
Here is how our documented features and policies apply to your questions:
Running agy on a remote server and supported regions (Questions 1 & 3):
Remote SSH support: The official, unmodified Antigravity CLI (v1.3.1) supports running on remote Linux environments and includes a built-in Remote SSH OAuth flow that provides a sign-in URL for your local browser and an authorization code to paste back into your remote SSH session.
Geographical availability: The United States, Australia, Japan, and South Korea are all on our supported countries list. Please note that account eligibility is based on the country associated with your account on the Google Terms of Service page.
Non-interactive scheduling vs. third-party coding agents (Questions 2 & 4):
Headless mode (agy -p): The CLI provides a documented Headless mode (agy -p "<prompt>" with --output-format json or stream-json) specifically designed to run non-interactively for your own scripts and CI pipelines using cached credentials. Standard headless CLI runs use your subscription’s quota and rate limits just like interactive sessions (Plans), as long as the setup is strictly for your own individual work, is not shared or exposed as a multi-user service, and does not run high-volume automated loops beyond normal individual use.
Third-party coding agents: Section 6 of the Terms of Service and our FAQ prohibit using third-party software, tools, or services to access or power Antigravity. We do not support using third-party coding agents (such as OpenCode, OpenClaw, Codex, or Claude Code) to orchestrate or wrap agy under a consumer login, as routing third-party agent prompts or automated loops through a consumer Antigravity login can trigger automated Terms of Service enforcement and account suspension.
Recommended route for unattended remote automation and OpenCode (Question 5):
Using OpenCode with a Gemini API key: Yes—if you would like to use a third-party coding agent such as OpenCode, using a Google AI Studio or Gemini Enterprise API key instead of your Antigravity login is the officially recommended approach (FAQ).
Using agy CLI with a Gemini API key: For unattended remote servers and CI runs where no browser is available, our CLI authentication guide also supports running agy directly with a Google AI Studio Gemini API key (by setting "modelProvider": "gemini" in ~/.gemini/antigravity-cli/settings.json and exporting GEMINI_API_KEY="your-api-key"), which sends model requests directly to the Gemini API without establishing a Google Account session.
Please let us know if you need any further information.