MCP server for SEO and GEO (Generative Engine Optimization) audits — on-page SEO, structured data, robots.txt, sitemaps, hreflang, and AI-crawler access for ChatGPT, Claude, Perplexity and Gemini. No API keys required.
Strong tool naming (verb_noun pattern, clear distinctions) and comprehensive descriptions (avg 180 chars, well above 10-char minimum). All 17 tools have explicit schemas with typed parameters and descriptions. Good composition: tools are single-responsibility and chainable (e.g., seo_audit includes geo option, geo_audit drills deeper). Main gaps: output schemas are not formally documented in code (inferred from descriptions), error handling lacks recovery guidance, and some parameter descriptions could be more prescriptive about constraints (e.g., URL format, domain format). No security issues detected (read-only tools, no secrets). STDIO transport caps protocol score but does not affect definition quality.
Check which AI crawlers (training vs citation) are allowed or blocked by robots.txt. Separates blocking a training crawler (GPTBot, ClaudeBot, CCBot) from blocking a citation crawler (OAI-SearchBot, Claude-SearchBot, PerplexityBot).
Check canonical host consistency: verify that www/non-www variants converge to a single canonical domain.
Analyze main content: word count, sentence count, paragraph count, Flesch reading ease, reading level, keyword density, and text-to-HTML ratio.
AI answer-engine readiness audit: checks AI crawler access, rendering, and llms.txt.
Analyze heading hierarchy (h1–h6), detect skipped levels, empty headings, and question-shaped headings (strong GEO signal).
Check hreflang tags: validate language codes, detect missing reciprocal links, and check for x-default.
Output schemas not formally documented in code. Tool descriptions imply structure (e.g., 'seven weighted sections', 'per-item success/failure') but JSON Schema definitions for response objects are not visible in source. LLMs cannot reliably parse unstructured output.
Error handling lacks recovery guidance. Tools return HTTP errors or parse failures but do not guide LLM on next steps (e.g., 'URL unreachable, try a different domain' or 'Invalid robots.txt syntax, check format at example.com/robots.txt'). Agents cannot self-correct.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
Check image SEO: alt text, title attributes, file sizes, dimensions, and lazy loading.
Audit internal and external links: check for broken links, redirect chains, and anchor text quality.
Check for llms.txt file at /.well-known/llms.txt. Note: llms.txt is a community proposal with no committed vendor support (Google has said it does not use it).
Check meta tags: title, description, robots, viewport, charset, and other on-page SEO signals.
Trace HTTP redirects: follow a URL chain and report each hop, final status, and redirect loops.
Check if a page renders without JavaScript. Most AI crawlers do not execute JavaScript, so a page can rank in Google and still be unreadable to every AI assistant.
Check robots.txt: parse rules, detect AI crawler blocks, list sitemaps, and validate syntax.
Composite SEO audit: one fetch, seven weighted sections, prioritised fixes. Pass include_geo=true to add AI answer-engine readiness in the same call.
Check sitemap.xml: parse URLs, detect duplicates, validate lastmod dates, and check for broken links.
Check Open Graph and Twitter Card tags for social media preview optimization.
Check JSON-LD structured data (schema.org), microdata, and RDFa for rich snippets and AI extraction.
Parameter constraints not fully specified. 'url' and 'site' parameters lack format validation hints (e.g., 'must be a valid HTTP(S) URL' or 'domain without protocol, e.g., example.com'). 'response_format' enum values are implied but not explicitly constrained in visible schema.
STDIO transport only. Server is not remotely accessible and cannot be used by hosted MCP clients (Claude.ai, Cursor, etc.). HTTP transport is available via environment variable but not the default. Limits production deployment options.
Tool composition could be optimized. 'seo_audit' with include_geo=true combines two concerns (SEO + GEO). While this is intentional for convenience, it prevents agents from calling seo_audit alone when GEO is not needed. Consider offering both seo_audit (SEO only) and seo_audit_with_geo (combined) as separate tools for flexibility.