An SEO and Generative Engine Optimization (GEO) suite for Google Antigravity, powered by Go.
Note
Lineage: Inspired by claude-seo by Daniel Agrici. Antigravity SEO replaces Python/Playwright runtimes with a single Go binary (seo-engine) and native agent tools ( search_web, read_url_content).
| Dimension | claude-seo + Claude Opus 5 | antigravity-seo + Gemini 3.8 Flash | Advantage |
|---|---|---|---|
| Input / Output Token Cost | $5.00 / $25.00 per 1M | $0.75 / $3.75 per 1M | 6.7x cheaper |
| Single Page Audit (~50k in / 10k out) | ~$0.50 | ~$0.075 | ~85% savings |
| Full Site Audit (20 pages, ~400k in / 80k out) | ~$4.00 – $8.00+ | ~$0.60 | ~85% – 90% savings |
| Multi-Agent Swarm Crawls | Not supported (single model tier) | ~$0.12 – $0.20 (viaflash_lite ) |
~25x – 40x cheaper |
| Live SERP & Crawl Data | Optional paid SaaS MCPs (DataForSEO, Firecrawl, Ahrefs) | ✅ Native search_web & Common Crawl |
Zero paid subscriptions |
| Engine Runtime | Python venv, pip dependencies, Playwright | ✅ Single Go binary (seo-engine) |
Zero dependencies, 10ms startup |
| Google Search Console | External Node/Python scripts | ✅ Native RSA JWT exchange | Built-in auth & queries |
Pricing based on September 2026 rates. Multi-agent crawls fan out to flash_lite subagents for penny-scale execution.
- Token Economics : Claude Opus 5 thinking traces bill at $25/1M, making multi-page audits costly ($4–$8+). Gemini 3.8 Flash ($0.75 / $3.75) cuts costs by 85–90% (~$0.60), while
flash_litesubagent swarms handle raw crawling for <$0.15. - Zero-Dependency Go Engine :
seo-engineis a standalone binary with 10ms execution, eliminating Python venvs, pip modules, and headless browser daemons. - Native Grounding : Live Google SERP analysis and markdown extraction use Antigravity's built-in
search_webandread_url_contentwithout paid third-party MCPs. - Built-in GSC & Backlinks : Direct RSA/PKCS8 JWT authentication for Google Search Console and keyless Common Crawl CDX index lookups.
- Native A4 PDF Reports : Built-in PDF report generation (
--pdf) viafpdfmatching executive print standards with zero external dependencies.
flowchart LR
User([User]) --> Skills["21 Antigravity Skills\nAudit • Schema • Content • GEO • Backlinks"]
Skills --> Engine["Go Engine (seo-engine)\nFast audits, crawler, GSC, & PDF generation"]
Skills --> Native["Native Grounding\nsearch_web & read_url_content"]
Engine --> Output["Deliverables\nCLI • JSON • Enterprise A4 PDF"]
macOS / Linux:
curl -fsSL https://raw.githubusercontent.com/angelotc/antigravity-seo/main/install.sh | bash
Windows (PowerShell):
irm https://raw.githubusercontent.com/angelotc/antigravity-seo/main/install.ps1 | iex
Installs the plugin to ~/.gemini/config/plugins/antigravity-seo and symlinks seo-engine into your PATH.
git clone https://github.com/angelotc/antigravity-seo.git
cd antigravity-seo
bash install.sh # On Windows: powershell -ExecutionPolicy Bypass -File install.ps1
Verify anytime with:
seo-engine doctor
| Command | Purpose |
|---|---|
headers <url> |
HTTP status, redirect chains, X-Robots-Tag, TTFB |
audit <url> |
On-page technical + schema audit |
page <url> |
Deep audit (technical + schema + images + content + hreflang) |
report <url> |
Executive HTML or native A4 PDF report ( --pdf ) |
schema <url> |
JSON-LD validation vs Google Rich Results (14+ types) |
images <url> |
Alt coverage, dimensions, CLS, formats, lazy |
content <url> |
Word count, headings, E-E-A-T signals, answer blocks |
hreflang <url> |
BCP47 codes, x-default, self-reference, duplicates |
llms <url> |
/llms.txt and/llms-full.txt AI discovery audit |
sitemap <url> |
Sitemap validation (or sitemap generate <url> ) |
robots <url> |
Robots.txt directives and AI crawler access policies |
drift <url> |
Change monitoring: baseline ,compare ,history |
backlinks <domain> |
Free Common Crawl link graph and domain captures |
gsc query|inspect |
Google Search Console Search Analytics and URL Inspection |
psi <url> /crux <url> |
PageSpeed Insights and Chrome UX Report field data |
indexnow <url...> |
Instant submission to Bing, Yandex, Seznam, and Naver |
All commands support --json. Core audit features are 100% keyless.
- Orchestration :
seo,seo-audit,seo-page - Technical & Content :
seo-technical,seo-content,seo-schema,seo-geo,seo-images,seo-hreflang,seo-drift,seo-backlinks - Search & Local :
seo-local,seo-maps,seo-ecommerce,seo-cluster,seo-sxo - Strategy :
seo-plan,seo-programmatic,seo-competitor-pages,seo-content-brief,seo-flow
The plugin also includes a PostToolUse schema hook that automatically validates JSON-LD files upon write, blocking deprecated schemas and placeholder text.
The engine and skills run 100% natively via standard shell execution (seo-engine <command> [options] <url> --json) with zero MCP middleware. Simply ensure seo-engine is in your $PATH (or run install.sh).
- Antigravity : Native execution via
run_command - OpenCode : Register skills path in
opencode.json:
"skills": {
"paths": ["~/.gemini/config/plugins/antigravity-seo/skills"]
}
- Claude Code : Symlink or copy skills into
~/.claude/skills/ - Codex / Cursor / Aider : Invoke
seo-enginedirectly via the terminal / bash tool
Upstream Parity vs claude-seo #
| Upstream Capability | Antigravity SEO Implementation |
|---|---|
| Runtime & Doctor | ✅ Pure Go runtime; zero Python venv |
| Audits (Technical, Schema, Content, Images, Hreflang) | ✅ Fast engine commands + 21 specialized skills |
| Sitemap Generator | ✅ Built-in crawler: sitemap generate |
| Schema Quality Gate | ✅ PostToolUse hook (bash + PowerShell) |
| Drift Tracking | ✅ Zero-dependency JSON snapshot store |
| Search Console (GSC) | ✅ Native RSA JWT service account exchange |
| Backlink Analysis | ✅ Free Common Crawl CDX index query |
| PDF Reporting | ✅ Pure Go A4 PDF engine ( fpdf ), zero external tools |
| SERP Grounding | ✅ Native Antigravity search_web (no paid MCP accounts needed) |
| SPA JavaScript Rendering | 🔁 Delegated to Antigravity's native browser tools |
MIT — see LICENSE.
source & further reading
github.com — original article
Show HN: Tokenhush – keeps your secrets out of what Claude Code sends
Show HN: Does per-step reasoning effort save money in Claude Code?
Check macOS Spotlight semantic indexing readiness