Kim OlsenAI consultant in Copenhagen. I build website AI agents for Nordic B2B companies.
Find out if your website turns away the buyer’s agent before it books.
More buyers ask Claude to read a pricing page and book the meeting for them. A bot rule written for scrapers can refuse that visit on the booking page, and nothing tells you. The checker reads your access log, matches every Claude-Agent request against Anthropic’s address list, and shows you which page said no.
What you get.
- The checker: one Python file, standard library only, that sorts every Claude request in your access log into real agent, fake, crawler and blocked
- An invented 34-line access log, robots.txt and bot rules for a five-page B2B site, so you can see it work before you run it on your own log
- A saved copy of Anthropic’s address list for offline runs, with a warning when it has expired
- A one-page note on what to allow, in English and Danish, to hand to whoever runs your firewall or CDN
- The 28 tests that try to fool it, and a README in English and Danish
What the sample log shows.
3 of 8
requests from real Claude agents were refused, all on the booking page, by a rule written for scrapers.
5 of 6
fake Claude-Agent requests, from addresses not on Anthropic’s list, were answered.
0
real requests refused when the same log is replayed with listed addresses let past the scraper rule. A replay, not a recorded result.
A check on an invented log.
The log, the site and the rules are invented, and Anthropic’s address list is the copy fetched on 8 November 2026. The checker trusts the address in your log: behind a CDN or proxy it may hold the proxy’s address, and the README says how to read the result then. It tells you what happened, it does not change your firewall.
Read what this means for your website and your buyers: A buyer’s agent tried to book a meeting. The scraper rule said no →