Adding knowledge sources

Path: Agent Studio → Knowledge → Sources

You build your agent’s knowledge one source at a time. Each source brings content in, keeps it in sync, and contributes the articles the agent can retrieve. You’ll usually combine several. Click + Add Source in the top-right to open the Add knowledge source picker, then pick the source that matches where the content already lives.

The source types

Crawl Website or Help Center

Point it at a starting URL and it indexes the whole public site, following links to pull in every page, then re-crawls on a schedule so published changes flow in on their own. Best when the canonical content is already live on your own domain. See Crawl a website or help center for setup and when to choose it over a single page or an API connector.

Add Single Web Page

Adds exactly one public page — a pricing page, a single FAQ, one blog post — without crawling everything around it. Use it when you want a specific page and nothing else, or to top up a crawl with a page that lives elsewhere.

Create Internal Article

Write a guide directly inside Foldspace, for knowledge that doesn’t exist publicly: SOPs, escalation steps, internal FAQs, or best-practice articles that shape how the agent runs an action (see the note below). Because you own the text, this is the cleanest source to edit when you’re closing a specific gap.

Help Center connector (Intercom, Zendesk)

Connects directly to your help-desk platform over its API instead of crawling HTML. For teams already on Intercom or Zendesk, this syncs your existing help-center articles automatically and keeps them current as your support team edits them, with no re-crawl and no copy-paste. Because it authenticates over the API, it can also reach help-center content that sits behind a customer login, which a crawl can’t.

Each platform has its own connection guide: Intercom authenticates with an access token, and Zendesk authenticates over OAuth. Both let you pick which collections or sections to sync and offer an optional weekly auto-resync.

Add Training Video

Uses an online video to train the agent and enrich answers with what’s shown on screen — product demos, tutorials, onboarding recordings. Good for walkthroughs where the steps are easier to show than to write.

Upload File (PDF, DOCX, or PPTX)

Imports content from a document — glossaries, pricing sheets, specs, slide decks. Use it for material that only exists as a file and isn’t going to live on the web.

Glossary

Defines key terms, abbreviations, brand names, and translations so the agent uses consistent vocabulary. Reach for it when your product’s language differs from how users phrase things.

Source	Syncs how	Best for
Crawl Website or Help Center	Scheduled re-crawl	Existing public KBs, docs, support sites
Add Single Web Page	Per-page refresh	One specific public page
Create Internal Article	You edit it directly	Non-public content; action best practices
Help Center connector (API)	Automatic, via the platform	Teams already on Intercom or Zendesk
Add Training Video	On the source	Demos and walkthroughs
Upload File	Re-upload	Glossaries, specs, decks
Glossary	You edit it directly	Terminology and translations

After you pick a type, provide the URL, file, or content as prompted. Each source shows its article count and sync status (for example, “Synced” with the date of the last update), and appears in the left sidebar under its category.

Knowledge powers actions too

The agent doesn’t only quote Knowledge in answers — it also uses it to operationalize actions. An internal article describing what makes a good Create Campaign (budget pacing, naming, required dates) is retrieved at runtime and applied when the agent fills and runs that action. Keep that kind of guidance in Knowledge; see Authoring actions.

Organizing with Folders

The Folders tab groups knowledge by topic or page context — useful when you want the agent to surface certain content only on certain pages of your app. Assign articles to a folder, then scope where that folder applies.

The knowledge article list

The main table lists every individual article with its Name, source URL (for web content), Location (folder or source), and Synced date. Use it to confirm a crawl or connector actually pulled in what you expected before you rely on it.

Next step

Once content is in, validate that the agent answers correctly from it before users do.