Claude-User
Reference: W67-HJM-87R
Fetches a page because a person asked an assistant. Vendor: anthropic.
Documented
What the vendor states. Each fact links to its source and the date we checked it.
| Fact | Value | Source |
|---|---|---|
| User agent, as documented | Not documented | |
| robots.txt behaviour | The vendor says it follows robots.txt | source, checked 2026-11-22 |
| Runs JavaScript | Not documented | |
| Published IP list | https://claude.com/crawling/bots.json | source, checked 2026-11-22 |
| Signs requests (Web Bot Auth key directory) | Not documented | |
| Reverse DNS suffixes | Not documented | |
| Network registered to the vendor | 160.79.104.0/21 | source, checked 2026-11-22 |
The vendor publishes one IP list for all of its bots. The list does not say which addresses belong to this agent.
Observed
Not observed yet.
Where documented and observed disagree
None found so far.
Prompted tests
- 2026-11-22 (our test site (probe)): We asked Claude (claude.ai) for the reference code on two unlinked, unindexed pages, one of them disallowed in robots.txt. Claude-User fetched robots.txt and the allowed page, and the answer quoted the correct code. It made no request for the disallowed page, and the answer said robots.txt blocks it. Source: Who reads the web probe findings, 2026-11-22.