Twitch Live Channel Scraper - Streams, Clips, VODs & Schedules

Scrape Twitch channels in one run: who is live right now and with how many viewers, plus their recent clips, recent VODs, published schedule and a channel search. Up to 100 channels per run, straight from Twitch's official Helix API.

Every row carries the same provenance fields whatever its type - when the event happened, when we observed it, a stable entity key and a content hash - so mixed output can still be joined, deduplicated and reasoned about later.

Record types

One run can return several types at once; the record_type field tells them apart.

record_type	one row per	event_time is
stream	channel that is live right now	when the stream started
clip	clip created in the lookback window	when the clip was created
video	VOD or highlight	when the video was created
schedule	scheduled segment	segment start time
channel	channel search result	when it went live, empty if offline

Fields on every row

field	meaning
record_type	stream, clip, video, schedule or channel
event_time	when this happened in the world (UTC ISO-8601), null when not applicable
ingest_time	when this run observed it (UTC ISO-8601)
entity_key	twitch:channel:<id>, twitch:clip:<id>, twitch:video:<id>, twitch:schedule:<channel>:<segment>
source_url	the page this record describes
source_id	the object's id at the source
content_hash	sha256 of the payload, so identical observations dedupe
http_status	status returned by the API for this record

Type-specific fields follow: viewer_count, game_name, title, language, view_count, duration_seconds, end_time, is_recurring and so on.

Input

{
  "channels": ["lofigirl", "gaules", "riotgames"],
  "includeStreams": true,
  "includeClips": true,
  "clipDays": 30,
  "maxClipsPerChannel": 20,
  "includeVideos": false,
  "includeSchedule": true,
  "searchQuery": "",
  "maxSearchResults": 20
}

Only the modes you enable cost anything. Channels that are offline produce no stream row - a run with zero rows is a valid answer, not a failure. Absence means "not live at ingest_time", which is information.

searchQuery works without any logins, for example "minecraft".

Output

{
  "record_type": "clip",
  "event_time": "2026-09-25T08:22:04Z",
  "ingest_time": "2026-10-23T13:40:11.902Z",
  "entity_key": "twitch:clip:InexpensiveVastPuppyBCouch-AaKwizGbhAYj_M6L",
  "source_id": "InexpensiveVastPuppyBCouch-AaKwizGbhAYj_M6L",
  "source_url": "https://clips.twitch.tv/InexpensiveVastPuppyBCouch-AaKwizGbhAYj_M6L",
  "content_hash": "51943e2b41cf...",
  "http_status": 200,
  "broadcaster_login": "gaules",
  "creator_name": "somebody",
  "title": "...",
  "view_count": 1204,
  "duration_seconds": 28.4,
  "language": "pt"
}

What this Actor cannot return

These need authorisation that belongs to the broadcaster, not to the Actor, so they are out of scope rather than missing:

data	why not
follower counts	moderator:read:followers on that channel, granted per broadcaster
subscriber counts	channel:read:subscriptions on that channel
chat messages	not in Helix at all; needs a persistent IRC connection

Anything claiming follower or subscriber history for arbitrary channels is either using a different, less reliable route or is out of date.

Pricing

Pay per result: $1.90 per 1,000 rows, plus a small per-run start fee. A row is a row whatever its type, so enabling clips and schedules costs more only because it returns more rows. Offline channels and empty modes cost nothing.

Notes

• Official Helix API. No website scraping, no proxies, no captchas, no bans.
• Requests stay well below Twitch's published rate limit (roughly 660/min against a budget of 800).
• VODs expire on Twitch after about 14 to 60 days, so video rows are recent by nature. Clips persist, which is what makes clip rows usable as history.