Wintro

Sources and limits

What Wintro uses

Wintro writes a brief about one person from public information it can find at the moment you ask, plus what you tell it. Every statement in a brief links to the page it came from, with that page’s date when the page shows one.

  • The open web, through search providers. Wintro searches for the person and their company using Serper, a search API, with Brave Search as a backup if Serper fails; the search words are the person’s name and company, plus the title or place you gave. It then reads the pages that match: company websites, news and press, local business journals, chamber of commerce and trade pages, award and recognition lists, personal sites, podcasts and interviews.
  • Public registries. The US National Provider Identifier (NPI) registry, for healthcare providers, using only: name, credential, practice address, enumeration date and specialty. OpenStreetMap business listings, for local businesses.
  • What you type. A name, a company, a title, a place, your own notes, and your “What I sell” line.
  • LinkedIn, only through you. Wintro never fetches LinkedIn from its servers. If you use the browser extension, it reads the profile you are looking at, in your browser, and hands it to your workspace once. LinkedIn’s terms restrict browser extensions that read profile data; the extension acts only when you click it, and any risk to your LinkedIn account is yours.
  • An AI model to read and write. Page excerpts and your inputs are sent to Anthropic’s Claude models to extract facts and write the brief. Dates, confidence levels, source labels and which claims are allowed to stand are decided by Wintro’s code, not by the model.

Sources the person or their company wrote about themselves (their own site, a sponsored article, a job posting) are shown, labelled “company-supplied” or “self-described”, and never counted as independent confirmation.

How Wintro behaves toward websites

Wintro identifies itself as WintroBot, follows each site’s robots.txt, and skips sites we have identified as forbidding it. It limits how fast it asks any one site for pages. If a site says no (an access-denied or blocked response), Wintro stops asking that site for a period of time, shows the page as “unavailable” in your brief, and does not try another way in.

For website owners

Wintro’s crawler identifies itself as WintroBot (every request carries a User-Agent beginning WintroBot/1.0, and that User-Agent includes the address of the WintroBot page, which describes the crawler). It reads public pages only: nothing behind a login, no forms, no workarounds.

  • robots.txt. WintroBot reads a site’s robots.txt before fetching a page and follows it. Rules written for WintroBot are used; if there are none, the rules for * apply. To keep WintroBot out entirely, add User-agent: WintroBot and Disallow: /.
  • Per-host limits. At most 2 requests at a time to any one site, with at least 0.3 seconds between request starts. On a “too many requests” or “unavailable” answer (429, 503) it backs off, honouring Retry-After, and gives up after 2 retries. When guessing where a company lists its people, it looks at no more than 6 pages of a site and stops after 3 “not found” answers in a row.
  • If a site says no. On an access-denied answer (403) or a refused connection, WintroBot stops requesting that site for 30 minutes and does not try another way in. It does not change its name or headers, rotate addresses, or retry around a block.
  • To ask us to stop. Email support@usewintro.com with the site’s address. We will add it to our do-not-fetch list and confirm. Blocking WintroBot in robots.txt also works and needs nothing from us.

What Wintro is built not to collect

  • Disciplinary, regulatory or complaint history. Wintro is built to exclude disciplinary, regulatory or complaint history from every source and drops pages that appear to be about it. The filter is automated and can miss; if you see something like this in a brief, flag it.
  • Sites that forbid automated access. Pages from sites we have identified as forbidding it are not fetched.
  • Data-broker and people-search sites. Contact-lookup, background-check and people-search sites we have identified are not used as sources: their pages are dropped before anything is fetched from them, including when another link redirects there. The list is kept by hand and can miss a site.
  • Contact details. Wintro does not look up, buy or list personal email addresses, phone numbers or home addresses. (A public page it reads may itself show one; Wintro does not search for them.) A practice or business address from a public registry or from a business’s own page may be shown, and can sometimes be a home office.
  • Registry fields beyond the list above. From the NPI file it never reads mailing addresses, phone or fax numbers, sex, tax IDs, licence numbers or other identifiers.
  • Private content. Nothing behind a login, nothing from private messages or private accounts.

What Wintro cannot guarantee

  • Accuracy. Names are shared by many people. Wintro checks that a page is about the right person and tells you how sure it is (high, medium, low), but it can be wrong, and pages themselves can be wrong or out of date. Check anything that matters before you rely on it. If a brief is about the wrong person, use “Wrong person?”; if a source is wrong, use “Not this person”.
  • Completeness. A brief shows what could be found. A short brief means little was found, not that nothing exists. Sources that are not switched on are listed as “not checked”.
  • Freshness. Sources are dated when the page shows a date; otherwise they are marked undated. “Recent” means within the last 12 months. Older facts are shown with their age.
  • Compliance of your outreach. Wintro is research, not legal advice. How you use a brief to contact someone (email and phone rules, privacy laws such as CAN-SPAM, TCPA, GDPR and CCPA, the terms of any platform you use) is your responsibility.
  • Third-party terms. Websites change their terms and technology without notice. Wintro blocks sources it knows forbid automated access, but it cannot promise that no page it reads will later be found to have restricted use.
  • Availability. Search and AI providers can fail or be rate-limited, and sites can refuse requests. A brief may be thinner, slower or fail; failed briefs and briefs that find no public record do not count toward your monthly limit.

Your data

Briefs live in your workspace until deleted. Shared research caches expire (search results after 7 days, research results after 30). Corrections you submit are stored with the brief context to improve Wintro. To have a person’s data deleted, or to have a person never researched again, contact support@usewintro.com; this removes the briefs, caches and discard records about them and blanks the identifying fields of any correction notes.