What it does
Sitemap Reader takes a domain (example.com) or the address of a sitemap and returns up to 200 page addresses with the date each one last changed (the sitemap's own lastmod, in UTC). If the file is a sitemap index it reads one level down: up to 5 child sitemaps on the same site. Given only a domain it reads /sitemap.xml, then the Sitemap: line of robots.txt. Every answer says what it read, what it skipped and whether the list was cut at 200, and carries a signed record (fingerprint of the list, counts, time) that anyone can check against the public key below. The addresses in the list are never visited. Nothing is stored, and no model is used.
- Who it is for
- Agents and people who need the page list of a site: to see what is on it, what changed lately, or what to read next (with Page to Text or Page Watch).
- Price
- Free, no sign-up: 40 reads an hour per address. Paid calls are planned and not charging.
- Limits
- Up to 200 addresses per call, in the order the file lists them, duplicates dropped; if there are more, the answer says so and is never silently cut. Public http(s) sites only, ports 80 and 443; private, local and metadata addresses are refused. At most 6 downloads per call, 1.5 MB each: the sitemap and up to 5 children. Child sitemaps must be on the same host (a different host, a .gz file or a sitemap inside a sitemap is skipped and counted). Compressed (.gz) sitemaps are not read, and neither are HTML, text or news-only feeds. A file with a DOCTYPE entity declaration is refused. Only the loc and lastmod of each entry are read; image, video and news extensions are ignored. A lastmod that cannot be read comes back as null, and a date without a time stays a date. Addresses that are not http/https, have a user name, or point to a private network are dropped and counted. A file that is cut off or broken part-way keeps the entries before the break and says it is incomplete. A site that blocks us or answers an error gives a clear error, not a guess. robots.txt is read only to find the sitemap; it is not used to decide what to skip. The signature shows LLPRO read these addresses from that file at that time. It does not show the pages exist, or that the dates are true.
For developers: try it
POST https://llpro.ai/mcp
{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"sitemap_read","arguments":{"site":"llpro.ai"}}}
-> urls (up to 200: url, lastmod), counts, what was skipped, signed recordFor agents
Tool sitemap_read on https://llpro.ai/mcp and in the tools list. Signed answers are checked with the public key below (signer llpro1:6ob6s4fvagwgwy5rammwtjp2di): an Ed25519 signature over the observation as canonical JSON (sorted keys, no spaces).
Public key (base64): vKPfwR7pD/7VswNmYoVU4pwvfSY0yJbrhUEoUQgknMs=
For agents: this page as Markdown · JSON · llms.txt · MCP endpoint llpro.ai/mcp · agent card