polite.aiDocs
BUILD / Knowledge

Website knowledge

A site's pages become answers agents can give — crawled, summarised, and kept fresh automatically.

The organisation's site is usually already there

The website address given at signup is crawled automatically the moment the account is approved — provided the crawl-consent box was ticked on the waitlist form. It becomes the organisation's first knowledge base, and it also gives the Builder a brief on the business, so new agents sound right from the first draft. Without consent, nothing is crawled: the base shows Consent withheld until the site is added manually.

Add another site

Open the form

Knowledge → Add knowledge, then the Website tab (the panel opens on whichever half the organisation is missing). A bare domain is fine — https:// is added automatically.

Confirm authorisation

Tick I'm authorised to have this site crawled. The button stays disabled until it is ticked.

Add website

The crawl starts straight away, with live progress on the list — see Knowledge bases.

One base per site: adding a host that already exists is refused. How many bases each plan includes is in Plans & limits.

The Website tab: address, the authorisation tick, and the login option for private sites.

What the crawler reads — and skips

The crawler reads the text of the site's pages — cleaned, summarised and indexed — and picks out contact details like emails and phone numbers as it goes. On public sites it honours robots.txt. Crawls are capped at a sensible size, and each base holds up to 100 MB of extracted text.

Password-protected sites

A knowledge base can sit behind a login — a staff area or an intranet. This site needs a login, ticked when adding the site, reveals the login type — Username & password form or HTTP Basic auth — and credential fields; form logins can also take a Login page address. If the login fails, the crawl fails with a check-your-credentials error on the row.

Good to know

The password is stored encrypted, used only by the crawler, and never shown again — the base simply carries a Private badge. Because it is the organisation's own consented content, authenticated crawls don't apply robots.txt.

Keeping it fresh

Website bases are re-crawled automatically on a schedule. After a site update that should reach agents now, Re-crawl on the row queues a fresh crawl immediately.

Crawl consent is revocable at any time. Deleting the base — Delete, then Confirm delete — withdraws it: everything stored from the crawl is purged permanently, and the site is not crawled again unless it is re-added.

Last updated 2026-08-18