Skip to content

Help center

Articles are markdown documents organized into collections (name, emoji, order). An article is draft until you publish it; only published articles are ever visible outside the dashboard.

Published content is served unauthenticated at:

GET /portal/{workspace_slug} → collections + article list
GET /portal/{workspace_slug}/articles/{article_slug} → one article (markdown body)

Drafts and unknown slugs return 404, so nothing unpublished can leak. The endpoints return JSON with open CORS and no framing restrictions — render them with your own front end or embed them where you need them.

Publishing an article syncs it into the auto-managed articles knowledge source, where it is chunked and embedded like any other document. That means:

  • The AI agent retrieves and cites published articles when answering.
  • Semantic search and the reply copilot see them.
  • Unpublishing or deleting an article removes its document from retrieval.

You never manage this source by hand — it mirrors the published state of your help center.

Once at least one article is published, the widget shows a Help tab automatically: visitors browse collections and read articles without leaving your site. The tab’s search runs over the embedded article chunks (the same retrieval pipeline as the agent), so it finds answers by meaning, not just title matches.