How do I use Google Antigravity for SEO and to get my site cited by AI?

Use Google Antigravity as the hands, not the judge: put your page rules in AGENTS.md or a glob rule in .agents/rules, connect an MCP server with serverUrl in mcp_config.json for live data and checks, and confirm each page on the published site before calling it done. Its screenshots show what the agent saw; they do not show indexing or AI visits.

Facts

Google announced Antigravity on November 20, 2025 as an agentic development platform whose agents plan, execute and verify tasks across the editor, terminal and browser. (Google Developers Blog: Build with Google Antigravity (Nov. 20, 2025))
Antigravity reads MCP servers from ~/.gemini/config/mcp_config.json for every project and from .agents/mcp_config.json in the workspace, as of 2026-10-13. (Google Antigravity docs: MCP)
Remote MCP servers in Antigravity need the serverUrl field; the url and httpUrl fields used by other agents are not supported, as of 2026-10-13. (Google Antigravity docs: MCP)
Rule files in .agents/rules take 1 of 4 triggers: always_on, model_decision, glob or manual. (Google Antigravity docs: Rules)
Antigravity truncates a rule file over 24,000 bytes, and global and always-on rules share a 20,000-token budget. (Google Antigravity docs: Rules)
Antigravity CLI, the agy command, reads GEMINI.md and AGENTS.md in the project and the same mcp_config.json files, as of 2026-10-13. (Google Antigravity docs: Migrating from Gemini CLI)

Does Google Antigravity do SEO by itself?

No, Antigravity does SEO through the rules you give it and the tools you connect. That is enough to write correct pages, but not to know whether Google indexed them, AI crawlers read them or AI assistants sent visits.

Where do I put SEO rules for Antigravity?

Put them in AGENTS.md at the repository root if other coding agents share it, or in a rule in .agents/rules with a glob trigger, so the rule loads only when Antigravity touches your page files and stays out of the shared always-on budget:

---
trigger: glob
description: SEO rules for new content pages
globs: app/**/page.tsx
---
- One H1 per page, phrased as the question the page answers
- Unique <title> and meta description; self-referencing canonical
- Content rendered on the server, not only after JavaScript runs
- JSON-LD only for what the page visibly says
- Add every new page to the sitemap
- Never change the home page or pages that already work

How do I connect an MCP server to Antigravity?

Open the agent panel's menu, choose MCP Servers, then Manage MCP Servers and View raw config, and add the server under serverUrl. For a server that signs in with OAuth, click Authenticate next to it in the Customizations tab:

// .agents/mcp_config.json (this project) or ~/.gemini/config/mcp_config.json (every project)
{
  "mcpServers": {
    "<name>": {
      "serverUrl": "<server-url>"
    }
  }
}

Antigravity checks its work in the browser. Is that enough?

No. Its screenshots and browser recordings are the agent's own evidence of what it did; these questions need the published page or outside data:

Question	Where the answer is
Does the page render with content without JavaScript?	Fetch the live URL, for example with curl
Are title, canonical and JSON-LD right on the published page?	The live HTML, not the local preview
Did Google index it?	Search Console
Did AI crawlers read it, did AI assistants send visits?	Server logs or an analytics tool that splits AI traffic

Step by step

Write the rules. Create AGENTS.md or a glob rule in .agents/rules with your page rules.
Connect the tools. Add the MCP server under serverUrl in mcp_config.json and authenticate it.
Ask for one batch of pages. Ask for a few new pages, each answering one question, without touching existing pages.
Review and deploy. Read the changes and the implementation plan, then publish the batch together.
Check live. Open each published URL, check content, title and canonical, then watch indexing and AI visits.

Common mistakes

Copying a config that uses url or httpUrl, which Antigravity does not support, instead of serverUrl.
Making a long SEO rule always_on, where it competes for the shared 20,000-token budget.
Treating the agent's screenshots as proof that pages are live and indexed.
Letting it rewrite the home page or pages that already bring traffic.

How easyIndex solves it

easyIndex connects to Antigravity as an MCP server under serverUrl, https://easyindex.app/api/mcp, and you authorize it in the browser. It is more than a server to call: the method comes inside it, with step-by-step instructions and the current rules, so there is nothing to install or keep updated. Antigravity makes every change, and easyIndex never touches your site or repository.

Each page is audited when submitted and checked on the live site after publishing, and a step completes only on evidence easyIndex checks, not on the agent's report or screenshots. The dashboard then shows AI-assistant visitors per page, which pages AI crawlers read and Search Console data, for US$ 30 per website a month after a 7-day free trial.

Frequently asked questions

Is Antigravity the same as Gemini CLI?

No. Antigravity CLI replaced Gemini CLI for many personal Google accounts in 2026; it reads GEMINI.md, but MCP servers move to mcp_config.json.

Can I use the same rules in Antigravity and Claude Code?

Yes, through AGENTS.md: Antigravity reads it as plain Markdown, always active for its folder.

Does the CLI use the same MCP servers as the app?

Yes. Antigravity CLI reads the same global and workspace mcp_config.json files.

Will Antigravity get my site cited by AI?

No tool can promise that. It can build clear, crawlable pages; each AI assistant decides what it cites.

Continue reading

Last verified on October 13, 2026.