Geoscore MCP Server

geoscore

Can ChatGPT, Claude and Perplexity actually see your site? geoscore checks a website's readiness for AI crawlers and search — its GEO (Generative Engine Optimization) — and hands back the fixes as copy-paste assets: a robots.txt that lets the named AI crawlers in, Schema.org Organization and FAQ markup, and an llms.txt. It runs locally over stdio, needs no account and no API key.

Five tools do the work. analyze_geo_score grades an existing domain from 0 to 100 with a per-category breakdown. get_geo_fixes renders the repair kit for that domain. make_site_geo_ready generates every asset for a brand-new site in one shot from a name, a URL and a description. suggest_monitoring_config drafts the monitoring setup for a business — target audiences, suggested queries, and the exact causabi.com API calls that would apply it. get_site_report returns a client-ready report: the full mention-rate breakdown when you have a causabi.com share slug, or the crawlability score on its own when you only have a domain.

Honest scope notes

• Labelled subset — get_product_guide is omitted. The original server ships six tools; the sixth returns the vendor's own guide document, whose contents are not published in any readable record, so an honest rebuild cannot reproduce it. Only the five tools above are built.
• The scoring weights are a labelled reconstruction. The original publishes its check list (crawlability; robots.txt per AI crawler; visible without JavaScript; Schema.org; FAQ; name and contacts; freshness / accessibility / llms.txt hygiene with a small weight) but not the weights. geoscore's composition — crawlability 20, robots 20, visibility 20, schema 15, FAQ 10, contacts 10, hygiene 5 — is this rebuild's, documented in src/geo.ts.
• The report wire path is a reconstruction. The share pages at causabi.com/report/<slug> are public; the API path behind them (https://causabi.com/api/v1/reports/<slug>) is this rebuild's rendering of that behaviour.

Configuration

None. No environment variables, no API key. Site facts are fetched directly from the domain being checked, with a fixed 15-second per-request timeout.

Tools

• analyze_geo_score — GEO score 0–100 plus a breakdown for an existing domain. The result is a JSON object: {domain, score, breakdown: {crawlability, robots, visibility, schema, faq, contacts, hygiene}, robots: {<crawler>: "allowed" | "blocked"}, facts: {home_status, has_llms_txt, has_json_ld}}. The eight crawlers tracked are GPTBot, ClaudeBot, PerplexityBot, ChatGPT-User, Claude-User, OAI-SearchBot, Google-Extended and CCBot. If the home page cannot be fetched at all the tool fails with site unreachable: <domain>; a reachable page with an HTTP error status is scored, not fatal.
• get_geo_fixes — the copy-paste fixes for a domain, as a markdown document with four sections: a robots.txt block allowing every named AI crawler (listing any the current robots.txt blocks), an Organization JSON-LD template filled with the site's title when discoverable, a FAQPage JSON-LD skeleton, and an llms.txt skeleton.
• make_site_geo_ready — all GEO assets for a new site in one shot (no API key needed). Takes name, url and description, plus optional services (array of strings), email, and faq (array of {question, answer}). Returns one document: Organization JSON-LD, a FAQPage JSON-LD block when faq is given, an llms.txt body (with a Services list when given), and a robots.txt allowing the named AI crawlers. Fully local — it never fetches.
• suggest_monitoring_config — draft a monitoring config from business_name and business_description, with an optional location. Returns JSON: {business_name, location, audiences, suggested_queries, apply_calls}, where apply_calls holds the two calls that apply the config on causabi.com — PUT https://causabi.com/api/v1/monitor/{site_id}/settings and PUT https://causabi.com/api/v1/monitor/{site_id}/queries — with the {site_id} placeholder. This tool itself makes no API call.
• get_site_report — pass report_slug for the full AI mention-rate breakdown (no API key needed), or just domain for a crawlability-score-only summary: the analyze_geo_score JSON plus a note field explaining how to get the full report. One of the two arguments is required.

CLI

npx @reneturcios/geoscore analyze_geo_score '{"domain":"example.com"}'
npx @reneturcios/geoscore get_geo_fixes '{"domain":"example.com"}'
npx @reneturcios/geoscore make_site_geo_ready '{"name":"Acme Corp","url":"https://acme.com","description":"We help small businesses grow."}'
npx @reneturcios/geoscore suggest_monitoring_config '{"business_name":"Bright Smiles Dental","business_description":"A family dental clinic.","location":"Austin, TX"}'
npx @reneturcios/geoscore get_site_report '{"report_slug":"abc123"}'
npx @reneturcios/geoscore --version

Each subcommand takes one JSON argument string and prints the tool's result text. Exit codes: 0 success, 1 tool error, 2 usage error.

MCP

npx -y @reneturcios/geoscore mcp

MCP client config:

{
  "mcpServers": {
    "geoscore": {
      "command": "npx",
      "args": ["-y", "@reneturcios/geoscore", "mcp"]
    }
  }
}

Docker

docker run -i ghcr.io/cloudcorprecords/geoscore          # stdio MCP server (CMD defaults to "mcp")
docker run --rm ghcr.io/cloudcorprecords/geoscore analyze_geo_score '{"domain":"example.com"}'   # CLI via container

Development

npm install
npm run build   # tsc -> dist/ (dist/cli.js is the bin)
npm test        # tsx --test test/*.test.ts — fixture-based, no network
npm run dev     # tsx src/cli.ts — run the CLI from source

License

MIT — see LICENSE (Copyright (c) 2026 Rene Turcios).