Sitemap Extractor (reliable) MCP server

Sitemap Extractor (reliable)

Give it a domain or a sitemap URL and get every URL the site publishes in its sitemaps, with lastmod, changefreq, priority, hreflang alternates and image counts. One bad sitemap never fails the run: it's logged as an error row, it costs nothing, and the rest of the site still comes through.

Why this one

Apify's own sitemap extractor requests every URL to check its status, on 4 GB of memory. In the 30 days to 11 Oct 2026, 130 of its 411 public runs failed, aborted or timed out. This Actor reads the sitemaps only, unless you ask for status checks, so a typical run finishes in seconds on 256 MB.

What it does

/sitemap.xml, /sitemap_index.xml, /sitemap.xml.gz and /wp-sitemap.xml.

• Finds the sitemaps for you. Reads Sitemap: lines in robots.txt, then tries

plain-text sitemaps, and XML with unescaped & that strict parsers reject.

• Follows sitemap indexes to any depth, de-duplicates sitemaps and URLs.
• Reads every format: XML urlsets, sitemap indexes, gzipped files (.gz or gzip bodies),

https:// (many robots.txt files still list the old scheme).

• Retries 429 and 5xx responses with backoff, and retries http:// sitemap links over

testing; a 3,000-URL news sitemap took under a second.

• Filters by include/exclude globs and by lastmod date, so you only pay for what you keep.
• Plain HTTP, no browser: fast and cheap. A 2,500-URL shop with 175 sitemap files took 39 s in

Input

{
  "startUrls": [{ "url": "https://www.allbirds.com" }],
  "includeGlobs": ["https://www.allbirds.com/products/*"],
  "maxUrls": 5000
}

Output

One dataset row per URL:

{
  "loc": "https://www.allbirds.com/products/womens-wool-runners-savanna-night-navy",
  "lastmod": "2026-10-11T22:35:07-07:00",
  "changefreq": "daily",
  "priority": null,
  "hreflang": [],
  "imageCount": 1,
  "sitemap": "https://www.allbirds.com/sitemap_products_1.xml?from=1878193471557&to=7369944137808",
  "error": null
}

A sitemap that can't be fetched or parsed shows up as a row with loc empty and error set (for example fetch failed: HTTPStatusError: Client error '403 Forbidden'). Run statistics are in the key-value store under STATS.

Price

Pay per result: $0.30 per 1,000 URLs (charged per url event; the best-known alternative charges $0.50 on the free plan). Error rows are free. With the default maxUrls of 1,000, a run never costs more than $0.30 unless you raise the cap.

Uses

SEO audits and content inventories, finding pages changed since a date, seeding a crawler or a RAG pipeline with a site's real URL list, monitoring a competitor's new product pages. Works well as an MCP tool for agents: one call, a clean list of URLs.

Ready-made examples (open one, change the domain, run):

Shortened. The full README is on GitHub.