Skip to main content
Your pepline agents are only as good as the knowledge they stand on. The pepline API is the inbound lane for that knowledge: push content from your CMS, sync scripts, no-code scenarios (Zapier, Make, n8n) or anything else that can send JSON — and bring in whole websites in one call.

​
The shape of it

  • One resource that matters: knowledge_sources — one per ORIGIN of knowledge your agents consult: a paste of text, a web page, a file, a whole website (its pages ride beneath as entries).
  • Sync-friendly by design: supply your own external_id and POST becomes an upsert. Re-send anytime; unchanged content short-circuits, changed content re-processes. Your system stays the source of truth.
  • Websites: point us at a site, we crawl its pages and keep each fresh as an entry under one source — honest page-level citations, per-page failure visibility, one refresh verb.
  • You own the trigger: there are no webhooks to configure (yet — poll for status; it typically settles in under a minute). You push when your data changes.

​
What happens to what you push

Every source is processed the same way regardless of how it arrived: pepline extracts the readable text, indexes it for the agents you activated it for, and writes a digest — a short statement of what the agent understood, which you can verify in the dashboard (Knowledge) or on the API object. Failures are never silent: a failed source carries a human-readable failure_reason.

​
Base URL

All requests are authenticated — see Authentication.