SEO MCP Server

This is a teaser tier, not a general public API: it has no key, no write actions, and intentionally modest fair-use limits. For a normal browser workflow, use the free tools directly.

Connect

Claude Code

claude mcp add --transport http patrickstox https://patrickstox.com/api/mcp

Claude Desktop or claude.ai custom connector

{
  "mcpServers": {
    "patrickstox": {
      "url": "https://patrickstox.com/api/mcp"
    }
  }
}

Cursor

{
  "mcp": {
    "servers": {
      "patrickstox": { "url": "https://patrickstox.com/api/mcp" }
    }
  }
}

Any Streamable HTTP MCP client can use the same endpoint: https://patrickstox.com/api/mcp. The server supports the stateless 2026-07-28 protocol and retains the 2025 initialization flow for older clients.

Requests are processed by our server and are not stored after processing. Anonymous run-level outcome counters may be used for aggregate research; URLs, domains, IPs, and identifiers are never included, and no statistic is released below 100 runs.

How to use it

1. Add the endpoint with the configuration for your MCP client.
2. Confirm the client discovers the tools below before relying on a prompt-driven call.
3. Use search_site for the complete public corpus or search_content for articles only, then fetch an article by its returned slug when full text is needed.
4. For verification tools, supply only the documented public IP, pasted markup, or public URL and preserve the returned evidence and error state.
5. Cite the canonical URL returned with published content.

Available tools

{
  "query": "in:technical-seo \"javascript seo\""
}

[
  {
    "slug": "javascript-seo",
    "title": "JavaScript SEO",
    "url": "https://patrickstox.com/technical-seo/javascript-seo/"
  }
]

{
  "query": "in:tools canonical"
}

[
  {
    "title": "Canonical Tag Checker",
    "kind": "tool",
    "url": "https://patrickstox.com/tools/canonical-checker/"
  }
]

{
  "slug": "javascript-seo"
}

{
  "content": [
    {
      "type": "text",
      "text": "# JavaScript SEO\n…"
    }
  ]
}

{}

[
  {
    "kind": "pillar",
    "label": "Technical SEO",
    "url": "https://patrickstox.com/technical-seo/",
    "pagerankShare": "8.42%"
  }
]

{
  "ip": "203.0.113.9",
  "claim": "googlebot"
}

{
  "verdict": "spoofed",
  "bot": {
    "id": "googlebot",
    "displayName": "Googlebot"
  },
  "findings": [
    {
      "findingId": "verify-bot.range-mismatch",
      "severity": "fail",
      "label": "Crawler IP is outside the published range"
    }
  ],
  "evidence": {
    "matchedRanges": [],
    "listsFetchedAt": "2026-07-05T16:45:33.112Z"
  }
}

{
  "input": "{\"@context\":\"https://schema.org\",\"@type\":\"Product\",\"name\":\"Example\"}"
}

{
  "entities": [
    {
      "type": "Product",
      "googleEligible": "does-not-meet"
    }
  ],
  "issues": [
    {
      "code": "google.product.offers-review-rating.missing",
      "tier": "blocks"
    }
  ]
}

{
  "url": "http://example.com/old"
}

{
  "hops": 2,
  "finalUrl": "https://example.com/new",
  "finalStatus": 200,
  "findings": [
    {
      "findingId": "trace-redirect.extra-hop",
      "severity": "warning",
      "label": "Redirect trace contains an unnecessary hop"
    }
  ],
  "chain": [
    {
      "status": 301,
      "kind": "http"
    },
    {
      "status": 302,
      "kind": "http"
    },
    {
      "status": 200,
      "kind": "http"
    }
  ]
}

{
  "url": "https://example.com/article",
  "ua": "facebookexternalhit"
}

{
  "finalUrl": "https://example.com/article",
  "httpStatus": 200,
  "userAgent": {
    "profile": "facebookexternalhit"
  },
  "acquisition": {
    "maxBytes": 262144,
    "capturedBytes": 18420,
    "truncated": false,
    "complete": true
  },
  "metadata": {
    "openGraph": {
      "og:title": "Example article",
      "og:image": "https://example.com/card.jpg"
    },
    "twitter": {}
  },
  "platforms": [
    {
      "id": "facebook",
      "title": {
        "value": "Example article",
        "source": "og:title"
      },
      "image": {
        "value": "https://example.com/card.jpg",
        "source": "og:image"
      }
    }
  ],
  "score": 97
}

Search syntax

The search tools use the same retrieval syntax as the site. Use in: for a section or taxonomy, type: for a result class, quotation marks for an exact phrase, and a leading minus sign to exclude a term. Multiple in: scopes are combined as alternatives. Available results still depend on each tool’s documented corpus.

• in:ai-search citations — search the AI Search pillar.
• type:guide "title tag" -generator — require the phrase in guides and exclude generators.
• in:tools canonical — search the Tools section with the full-site search tool.

Example project Static documented response

The search_content example above scopes an exact javascript seo phrase to Technical SEO and returns a result shaped like this:

This is a deliberately trimmed documentation fixture from the shared tool roster, not a live request made when the page renders.

What each step produces

Content tools return source records intended to retain their canonical URLs. verify_bot returns identity evidence and a crawler-range snapshot date. validate_schema returns parsed entities and issue tiers. trace_redirect_chain returns a bounded chain and final state. inspect_og returns normalized Open Graph and Twitter fields, resolved image URLs, selected-UA evidence, and per-platform source provenance. Every verification result also carries a versioned capture-provenance object and finding packet so an agent can preserve the observed mode, JavaScript state, coverage, evidence, fix, and verification instruction. Neither URL tool exposes full HTML or arbitrary response headers. Treat an error, truncated capture, or unavailable check as not evaluated, never as a negative verdict.

How project data is stored

The page and Worker share the same MCP registry and free-tier availability rule, so names, argument schemas, rate tiers, and documentation examples stay aligned with the tools an unauthenticated client can list. Modern clients use stateless per-request metadata and server/discover; legacy clients can still negotiate through initialize. Streamable HTTP requests are routed to read-only handlers. Content calls use the build-time published corpus; tool calls reuse bounded validators and safe public-network checks.

Features

• Search the complete public site corpus or only published SEO guidance, with canonical attribution.
• Fetch full article markdown and list the public content taxonomy with build-time PageRank share.
• Verify one crawler IP, validate pasted schema, trace one public redirect chain, or inspect social metadata from one public page.
• Return versioned capture provenance and portable finding packets with verification-tool results.
• Supports MCP 2026-07-28 plus the legacy 2025 initialization flow over Streamable HTTP, without a key.

Limits & fair use

• Published-content tools: 120 calls per minute per IP.
• Tool calls (bot verification, schema validation, redirect tracing, and social metadata inspection): 10 calls per day per IP across the teaser tool set.
• URL inspection accepts only public URLs and allowlisted user agents, validates every redirect hop, and bounds redirects, bytes, and time.

Limits protect the free service. A paid API tier is not available yet.

Common request error: Provide a public http(s) URL. Correct the request arguments or wait for the stated rate-limit window before retrying. Call the same read-only tool again and confirm it returns a result object instead of an MCP error.

For an unexpected connection or service error, check the current service status before retrying or reporting it.

Limitations

• No write actions, private content, Search Console access, bulk crawling, or general proxy behavior.
• Public tool calls share a small daily per-IP teaser quota.
• Bot verification depends on dated published ranges and available DNS evidence.
• Schema checks are deterministic eligibility guidance, not a guarantee of a rich result.
• URL checks are bounded and accept only public HTTP(S) targets.
• A truncated inspect_og capture may report observed tags, but missing fields, completeness, and score remain explicitly not evaluated.

Attribution

When you use material returned by get_article, cite its canonical Patrick Stox URL. The server flags AI-assisted drafts pending human review; treat that status as part of the source context.

Frequently asked questions