/v1, for everything the
platform does: monitor sources (YouTube channels, videos, RSS, X), transcribe
and analyze them with agents, organize them into collections, and search over the
whole library.
The model
Items
A monitored source — a YouTube channel, a single video, an RSS feed. The
atomic unit; everything else points back to items.
Collections
Groups of items. Private or public; can be paid (fans subscribe).
Reporters
Agents that process sources. A reporter runs workflows (ordered
steps, each a tool + prompt) and can run on a schedule.
Executions
Every run (transcript, workflow, schedule) creates an execution you can
track to completion.
Transcripts & documents
Artifacts produced from items: transcripts (one per language) and documents
(summaries and other workflow output).
Search & discover
Semantic search over your library, plus a public catalog (
discover) of
collections, reporters and users.Base URL
The whole API lives under/v1:
Two credentials, one API
The same endpoints accept either credential — only the billing differs:- API key (
ck_live_…) — for integrators. Drawn per call from the key’s balance (flat platform fee + pass-through). This is what these guides use. - Session JWT — what the web app uses (from
POST /v1/auth/login). No platform fee, only pass-through.
Next
Make your first call in the quickstart, then read
Core concepts — especially asynchronous operations,
which most write endpoints rely on.