ToraSEO MCP server - exposes SEO tools to Claude and other MCP-compatible AI clients
ToraSEO MCP exposes 11 SEO analysis tools via STDIO. All tools have basic descriptions (50-150 chars) and single-parameter schemas (url: string). However, critical gaps undermine production readiness: (1) Parameter descriptions are minimal or absent, most just repeat 'The URL to [action]' without explaining expected format, constraints, or error cases. (2) Output schemas are not documented in the visible code, LLMs cannot predict return structure or plan downstream calls. (3) No error handling guidance, tools fail silently or return raw HTTP errors. (4) Tool names lack clarity on scope: 'analyze_*' tools overlap conceptually (analyze_meta, analyze_canonical, analyze_headings all inspect page structure). (5) No pagination, batching, or result limits documented. Baseline: 549 production tools average 194-char descriptions with 100% param coverage; ToraSEO averages ~80 chars with ~10% param descriptions. This is a functional but immature implementation.
Checks canonical as a separate on-page block: missing, relative, or pointing elsewhere.
Extracts main text and counts words, links, images, and text-to-code ratio.
Checks H1-H6, skipped levels, empty headings, and page outline structure.
Checks robots.txt and meta robots as local indexability signals. Does not claim live search-index presence.
Counts internal, external, and invalid links in the extracted main content.
Checks title, description, canonical, robots, Open Graph, Twitter Card, and technical meta tags.
Finds the sitemap and checks basic sitemap structure issues.
Output schemas not documented. LLMs cannot predict return structure (fields, types, nesting) for any of the 11 tools. This forces agents to guess at response shape and plan blind.
Parameter descriptions missing or trivial. 'The URL to scan' repeated across all 11 tools. No guidance on format (http/https required?), constraints (max length?), or error cases (invalid domain, timeout).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
Checks redirect chains, loops, failures, and excessive hops.
Checks whether ToraSEO may scan the page according to robots.txt rules.
Detects public CMS, framework, analytics, CDN, and server signals.
A single-shot probe that checks basic SEO signals: status, title, H1, description, and response time.
No error handling guidance. Tools fail silently or return raw HTTP errors. LLMs receive no recovery hints (e.g., 'URL unreachable, try a different domain' or 'robots.txt not found, site may allow all crawlers').
Overlapping tool semantics. 'analyze_meta', 'analyze_canonical', 'analyze_headings' all inspect on-page structure. LLMs conflate them. Consider consolidating into 'analyze_page_structure' with a 'component' enum (meta|canonical|headings|all).
No pagination or result limits documented. Tools like 'analyze_links' and 'analyze_content' may return large datasets. No guidance on max results, pagination tokens, or truncation behavior.