Websites
Set up RSS / Atom feeds
Add news and blog feeds — industry news, competitor blogs, regulator announcements, your own release notes — and every entry becomes a dated document tied to its feed's name, so the assistant also knows what's being published around you. Read-only, no credentials.
Before you start
- No credential — any public feed address works; the feed's site only ever sees an ordinary request, and nothing is written anywhere.
- Collect the feed addresses: most sites link theirs in the page header or at /feed, /rss or /atom.xml; a browser extension or an “RSS” link on the site usually reveals it.
- Only public addresses are fetched — an address on a private or internal network is refused.
- Documents → RSS / Atom feeds, paste the addresses into Feed URLs, one per line (https:// is assumed if you leave it off).
- Choose access groups, Test, then Sync.
What comes in
- Every entry in every listed feed becomes one dated document headed with its date, its link and its feed's name, carrying the entry's full content when the feed includes it and its summary otherwise, HTML stripped — RSS 2.0, RSS 1.0 and Atom all work.
- Feeds only ever show a recent window: entries already synced stay searchable after they scroll out of the feed, and a feed you later remove from the list leaves what it brought in searchable too.
- An entry with no title and no body is skipped; one with no stable id can duplicate if it's later edited at the source, and Sync notes it when a feed has such entries.
- One broken feed never fails the others: it's reported in the sync notes and its earlier entries stay; the sync fails only when none of the feeds could be read.
Access
Access is allow-only: every entry this source brings in is visible to the groups you choose when setting it up — a public feed carries no reader list of its own.
Feed content ranks low — the outside word, never a company record — so it never outranks your own documents in an answer. Up to 50 feeds and 500 entries per feed are read each sync; Sync notes it when either cap is hit. Test fetches only the first listed feed to confirm the address and network path both work, names it and counts its entries; the rest sync unprobed. A site that blocks the request fails with a refused-to-serve reason — usually its bot protection, so allow this deployment's address if the site is yours — and a page that isn't a feed is reported as not a recognisable RSS or Atom feed.
Last updated 20 Sep 2026