Privacy
This service is built for agents, and it collects very little — there are no accounts, no advertising, no tracking cookies, and nothing here is sold to anyone.
What we log
Queries: search terms and filters sent to discovery and resolve endpoints, with a hashed client identifier and coarse metadata — used for rate-limiting, abuse control, and the public demand feed at /state/demand. Submissions: listing metadata plus an optional submitter email, used for status notifications and disputes. Correction: from launch on 2026-07-19 until 2026-09-01, submitter emails were also exposed on public listing-detail reads — 94 listings were affected; see the eleventh correction. Payer reports: transaction hashes, wallet addresses, and signatures — these are retained and publicly re-verifiable by design; wallet addresses and transactions referenced are already public on-chain. Operational logs: IP addresses appear transiently in rate-limit counters and abuse controls, not in published data.
Search terms, specifically (updated 2026-09-11)
The demand feed publishes the search terms themselves, verbatim, with counts of how many searches and how many distinct client IPs asked them. It never publishes the client identifier, the IP, or which client asked what. Three rules bound it, and they apply only to searches made from 2026-09-11 onward — terms logged before that date stay under the stricter rule they were collected under (three distinct clients minimum) and are not republished under the new one.
- Never published: a term containing a URL, a hostname, a wallet address, or an email address. These identify a party rather than describe a need, and no count makes them safe.
- Amended 2026-09-12: one searcher is enough. The rule here required two independent clients before a term appeared. In practice it withheld 46 of 47 terms — agents describe the same need in different words, so almost nothing is asked twice verbatim — and what it withheld were generic category phrases ("calendar holidays", "email verification", "flight travel") that identify nobody. From 2026-09-12 09:54 UTC a term is published however many clients asked it; the protection is the content rule above plus the withhold list below. Terms logged between 2026-09-11 15:00 and this amendment were collected under the two-searcher rule and are not republished under this one: the feed's window begins at this amendment. (Wording corrected 2026-09-24, correction #25: it said those terms “stay under the two-searcher rule”, which read as if they were still published under it. They are not published at all.)
- Withheld by hand, and counted. Any term can be removed from the feed — on request, or on our own reading that it reveals something specific about an identifiable party rather than describing a need. The number withheld this way is published alongside the feed, so the removals are visible even though the terms are not.
- Sellers checking their own listings are excluded from the signal and counted separately, so the exclusion is visible rather than silent.
If you searched for something you would rather was not published, write to hello@nohumans.directory and we will remove it from the feed.
What we publish
Listing metadata, verification results, probe statistics, aggregated payer-report outcomes, and the query statistics described above. Nothing published identifies an individual person — with one exception, now fixed: submitter emails were published on detail reads from launch on 2026-07-19 until 2026-09-01 (eleventh correction). Wallet addresses attached to reports are pseudonymous and were already public on-chain.
Your data
To correct or remove a submitter email, delist a listing you control, or ask anything about data we hold, write to hello@nohumans.directory. On-chain data referenced by reports cannot be deleted from the chain by anyone, including us.
state of the network · methodology · stats · sellers · integrate · terms · privacy