The SEO Crawler MCP is a self-contained, local website crawling and SEO analysis server backed by SQLite storage. It provides four core tools:
run_seo_audit — Crawl any website (up to 10,000 pages, 10 levels deep) using Chrome or Googlebot user agents, storing page data, links, headers, and metadata in a local SQLite database.
analyze_seo — Run 25+ SQL queries across five categories, returning structured reports with affected URLs and fix recommendations in structured, summary, or detailed formats.
query_seo_data — Execute individual named queries (e.g., missing-titles, broken-internal-links, orphan-pages, missing-csp) for targeted results.
list_seo_queries — Discover all 28 predefined queries, filterable by category or priority (CRITICAL/HIGH/MEDIUM/LOW).
What gets detected:
Critical: Missing titles, broken internal links, 4xx/5xx errors
Content: Duplicate/missing titles, descriptions, H1s, thin content
Technical: Redirect chains, orphan pages, canonical mismatches, non-HTTPS pages
Security: Missing CSP/HSTS/X-Frame-Options/Referrer-Policy headers, unsafe external links
Optimisation: Title/description length issues, missing images, excessive external links
Everything runs locally with no external API dependencies. It supports both AI assistant orchestration via MCP tools and direct terminal CLI usage for large or scheduled crawls.
Provides an embedded SQLite database for storing website crawl data, allowing AI agents to perform automated SEO audits and custom SQL queries to analyze site structure, metadata, and technical issues.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@SEO Crawler MCP crawl https://example.com and audit it for technical SEO issues"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
This project has been superseded by SEO Audit Console.
SEO Crawler crawled your whole site into SQLite and gave Claude 25+ analysis queries over the result. The crawler survived the move - it was rewritten to be stdio-safe and it is now the crawl half of a bigger tool.
SEO Audit Console keeps the crawl and adds the thing the crawl could never tell you on its own: what people actually searched for to get there. It merges the crawl with your Google Search Console history and on-demand DataForSEO data, then ranks every finding by the clicks it could recover rather than by severity.
A flat crawler sells you severity. This ranks by yield.
git clone https://github.com/houtini-ai/seo-audit.gitDocs and setup: github.com/houtini-ai/seo-audit
Related MCP server: seoagent
The npm package still installs. @houtini/seo-crawler-mcp is deprecated, not unpublished. Nothing you have running will break.
This repository is archived. The code stays readable and every existing link keeps working, but there are no further releases, and issues and pull requests are closed.
The crawler got better on the way over. It no longer depends on Crawlee (which logged to stdout and corrupted MCP's JSON-RPC framing), it seeds discovery from sitemaps and known GSC URLs rather than links alone, and it guards against redirects that wander off-host.
Start a crawl on the new tool and it builds its own database:
start_crawl for https://example.com/Run refresh_property instead if you want the crawl, the Search Console sync and URL inspection in one pass.
Part of the Houtini open-source MCP set. Questions: hello@houtini.com.
Analyze SEO data from a completed crawl. Runs 25+ SQL queries to detect critical issues, content problems, technical SEO issues, security vulnerabilities, and optimization opportunities. Returns structured report with affected URLs and fix recommendations.
Inputs: crawlPath (required), includeCategories, maxExamplesPerIssue, format.
List all available SEO analysis queries with descriptions, priorities, and fix recommendations. Optionally filter by category or priority level.
Inputs: category, priority.
Execute a specific SEO analysis query by name. Use list_seo_queries to see available queries. Returns detailed results with affected URLs and context.
Inputs: crawlPath (required), query (required), limit.
Crawl a website and extract comprehensive SEO data using Crawlee HttpCrawler. Returns crawl ID and output path.
Inputs: url (required), maxPages, depth, userAgent.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 4 tools
The tools have mostly distinct purposes, but there is some overlap between analyze_seo and query_seo_data, as both involve executing SEO analysis queries. However, analyze_seo runs a predefined set of queries, while query_seo_data allows for executing specific queries by name, which helps differentiate them. The other tools (list_seo_queries and run_seo_audit) are clearly distinct.
All tool names follow a consistent verb_noun pattern with snake_case (e.g., analyze_seo, list_seo_queries, query_seo_data, run_seo_audit). The naming is predictable and readable, with no deviations in style or convention.
With 4 tools, the count is reasonable for an SEO crawler server, covering key operations like crawling, listing queries, executing queries, and analyzing data. It might be slightly thin for a comprehensive SEO toolset, but it is well-scoped and each tool earns its place.
The toolset covers core SEO analysis workflows, including crawling, query listing, and data analysis. However, there are notable gaps, such as missing update or delete operations for managing queries or crawl data, and no tools for monitoring or reporting beyond the initial analysis. This could limit agent flexibility in handling ongoing SEO tasks.
Crawls a website for broken links, redirects and SEO issues, and lists changes since the last run
Website health, SEO, security-header, SSL and broken-link audits for AI agents.
Technical SEO audit crawler: broken links, 404s, redirects, titles, canonicals. Runs on Apify.
Crawl a site for broken links, 404s, dead images, redirect chains and slow pages, with sources