Knowledge in the dashboard sidebar lists every knowledge base the organisation owns — each one a collection of content agents can search and answer from on live calls.
Two kinds of base
- Websites. The site is crawled, every page cleaned and summarised, and the result indexed. The organisation's own site is usually there already — crawled automatically when the account is approved. See Website knowledge.
- PDF documents. Manuals, price lists or policies upload as a named collection. Add PDFs on the row adds more later; re-uploading a file with the same name replaces it. Files can be up to 50 MB each; scanned image-only PDFs aren't supported yet.
Each row shows the base's name, a kind badge, a one-line detail — page or document counts, contacts found, size used — the last-crawled date, and a status badge: Queued, Crawling or Processing (with a live page count), Ready (with totals), Failed (with the reason), or Consent withheld (crawl consent was not given at signup, so the site has been left alone).
How agents answer from knowledge
At call time, a knowledge-enabled agent looks things up live: it searches the indexed pages and documents, pulls the full text of the best matches, and can fetch the company's contact details — answers are grounded in what the content actually says.
Which agents get access is decided in the Builder — a request naming which agents should answer from the website or documents is enough. Some voice agent types can't use the knowledge tool directly; for those, the Builder quietly adds a text helper agent, visible on the team graph.
The emergency switch
Once the first content is ready, a single toggle sits at the top of the list: Knowledge base is live. Switched off, every agent in the organisation immediately loses the knowledge tool — lookups fail, calls otherwise carry on. It exists for the moment something wrong has been crawled and needs to be out of callers' ears right now.
Plan allowances
The plan sets how many bases an organisation can have — 1 on pay as you go, 2 on Pro, 10 on Scale — and the header shows the position ("1 of 2 used"). Every base holds up to 100 MB of extracted text. After a downgrade below current usage, existing bases keep working, but no more can be added — see Plans & limits.
Deleting a base — Delete, then Confirm delete — purges its content permanently and frees a slot straight away.
Anyone in the organisation can view this screen; adding, re-crawling, deleting and the emergency switch are for owners and admins — see Team & roles.