mcp-seo-audit

On-page SEO analysis as MCP tools.

An MCP server that exposes on-page SEO analysis — metas, structured data, robots.txt, sitemaps and links — as tools any MCP-compatible client can call. The six core tools need no API keys and no accounts: everything runs off plain public-page fetches. check_vitals optionally adds real-user Core Web Vitals with a free API key.

Tools

| Tool | Returns | |---|---| | audit_page | Title/meta lengths, canonical, robots meta, Open Graph & Twitter cards, headings outline, image alt coverage, word count, lang, hreflang | | extract_schema | Every JSON-LD block parsed, @type values, parse errors, and validation against Google's rich result requirements (missing required vs. recommended properties) | | check_robots | robots.txt user-agent groups, allow/disallow rules, declared sitemaps | | parse_sitemap | URL counts, lastmod sample, nested sitemap indexes | | extract_links | Internal/external/nofollow split, optional broken-link check | | crawl_site | Breadth-first crawl of internal links, auditing every page and aggregating findings by issue category | | check_vitals | Real-user Core Web Vitals from the Chrome UX Report: LCP, CLS, INP (plus FCP and TTFB) at the 75th percentile with Google's good / needs improvement / poor verdicts — needs a free API key |

The server returns structured data and leaves the interpretation to the client. The same tools drive a quick audit, a competitor comparison, or a full crawl conversation.

Install

Requires Node.js 20.18.1 or newer — cheerio pulls in undici 7, which needs the File global introduced in Node 20. On older versions the server fails to start.

No install step — register the server with any MCP client and npx fetches it on first run:

{
  "mcpServers": {
    "seo-audit": {
      "command": "npx",
      "args": ["mcp-seo-audit"]
    }
  }
}

On Windows, wrap the command in cmd:

{
  "mcpServers": {
    "seo-audit": {
      "command": "cmd",
      "args": ["/c", "npx", "mcp-seo-audit"]
    }
  }
}

To try it outside a client, npx mcp-seo-audit prints mcp-seo-audit ready and waits for a client on stdio.

From source

For contributors, or to run an unreleased change:

git clone https://github.com/mk-techi/mcp-seo-audit
cd mcp-seo-audit
npm install
npm run dev

Then point the client at the local checkout instead of the published package:

{
  "mcpServers": {
    "seo-audit": {
      "command": "npx",
      "args": ["tsx", "/absolute/path/to/mcp-seo-audit/src/index.ts"]
    }
  }
}

Core Web Vitals (CrUX) API key

check_vitals queries Google's Chrome UX Report API — the Core Web Vitals real Chrome users experienced over the last 28 days, the same field data Google Search uses. It reads a key from the CRUX_API_KEY environment variable:

{
  "mcpServers": {
    "seo-audit": {
      "command": "npx",
      "args": ["mcp-seo-audit"],
      "env": { "CRUX_API_KEY": "your-key" }
    }
  }
}

The key is free and needs no billing account: create an API key in the Google Cloud console and enable the Chrome UX Report API for its project. Without the key, check_vitals explains how to get one and every other tool keeps working.

Usage

Once connected, the tools are available. Example prompts:

• Audit example.com and give me a prioritized fix list.
• Does this page have valid JSON-LD? Which types, and what's missing for rich results?
• Compare the on-page SEO of my landing page against a competitor's.
• Find broken internal links on the homepage.
• How are my real-user Core Web Vitals on mobile, and which metric should I fix first?

Build

npm run build   # emits dist/
npm start       # runs the compiled server

Roadmap

• [x] crawl_site - follow internal links up to N pages and aggregate issues
• [x] Core Web Vitals via the public CrUX API
• [x] Schema validation against Google's rich results requirements
• [x] Publish to npm (npx mcp-seo-audit)

License

MIT