Twitch Scraper

Scrape Twitch — top live streams, channel info, recent VODs, top clips, and game (category) directories. Pure HTTP via Twitch's public GraphQL endpoint. No login or API key required.

What this actor does

Twitch's public website is powered by a GraphQL API at gql.twitch.tv. This actor talks to that API directly using the public web Client-ID, the same identifier the Twitch web frontend uses for unauthenticated browsing. That means the actor can pull:

• The current top live streams across Twitch (filterable by language)
• The top games/categories on Twitch right now
• Live streams within a single category
• Channel profile + recent VODs + top clips for any channel
• Top clips for a specific category over a chosen time window
• Single VODs by video ID
• Single clips by slug
• Full search across channels, categories, and videos
• Anything you can identify by a Twitch URL (channel, video, clip, or category page)

All output records are flat JSON with snake-case-style camelCase keys, omit-empty (no null/""/[]/{} anywhere in the dataset).

Modes

Mode	What you get
Top live streams (topStreams)	The most-viewed live streams across Twitch right now, with broadcaster, game, viewer count, language, tags, and preview thumbnail. Optional language filter.
Top games / categories (topGames)	The top categories on Twitch right now — sorted by current viewer count — with box art, follower counts, and broadcaster counts.
Search Twitch (search)	Free-text search across channels (users), games (categories), and videos. Restrict by searchType to channel, game, or video only.
Channel info + VODs + clips (byChannel)	Channel profile (followers, partner status, current live state, last broadcast, banner / profile / offline images) plus recent VODs and top clips.
Browse a category (byCategory)	Live streams currently broadcasting in a given category — filterable by language.
Top clips for a category (topClips)	Top clips for a category over a configurable time window (last day / week / month / all time).
Lookup VOD by ID (byVideo)	Full metadata for a single VOD — title, length, views, owner, game, broadcast type, language.
Lookup clip by slug (byClip)	Full metadata for a single clip — broadcaster, curator, source video, view count, embed URL.
Resolve a Twitch URL (byUrl)	Paste any list of Twitch URLs and the actor resolves each one to the right query (channel / VOD / clip / category page).

Filters

All filters apply across modes wherever the field exists. Records that don't have the field pass through (filter is opt-in, not opt-out).

• Broadcast language — restrict live streams and search to EN / ES / DE / FR / PT / RU / JA / KO / ZH / IT / TR / PL / AR / TH / ID / VI / NL / SV / FI / DA / NO / CS / HU / EL / RO / BG / UK / HE / MS / TL / ASL / OTHER (or (any) for no filter).
• Min concurrent viewers — drop live streams below this count.
• Min total views — drop VODs / clips below this view count.
• Title contains keyword — case-insensitive substring match against stream / video / clip titles.
• VOD type (byChannel) — ARCHIVE, HIGHLIGHT, UPLOAD, PAST_PREMIERE, PREMIERE_UPLOAD.
• Video sort (byChannel) — TIME (most recent first) or VIEWS (most-viewed first).
• Clip time window — LAST_DAY, LAST_WEEK, LAST_MONTH, or ALL_TIME.

Output fields (per record)

Every record includes a recordType discriminator (stream, channel, video, clip, game, or error) and a scrapedAt ISO-8601 timestamp.

Live stream (recordType: "stream")

streamId, title, viewersCount, streamType, language, startedAt, previewImageURL, channelId, channelName, displayName, gameId, gameName, gameSlug, tags, broadcaster (sub-object: id/login/displayName/profileImageURL/primaryColorHex), game (sub-object), url.

Channel (recordType: "channel")

channelId, channelName, displayName, description, createdAt, profileImageURL, bannerImageURL, offlineImageURL, primaryColorHex, isPartner, isAffiliate, isStaff, followersCount, streamTitle, currentGame (sub-object), liveStream (sub-object — only present when live), isLive, lastBroadcast (sub-object), url.

VOD (recordType: "video")

videoId, title, description, lengthSeconds, viewCount, createdAt, publishedAt, previewImageURL, broadcastType, language, gameId, gameName, gameSlug, game (sub-object), channelId, channelName, displayName, owner (sub-object), url.

Clip (recordType: "clip")

clipId, slug, title, viewCount, durationSeconds, createdAt, language, thumbnailURL, embedURL, videoOffsetSeconds, gameId, gameName, gameSlug, game (sub-object), channelId, channelName, displayName, broadcaster (sub-object), curator (sub-object), sourceVideo (sub-object), url.

Game / category (recordType: "game")

gameId, name, displayName, slug, viewersCount, followersCount, broadcastersCount, boxArtURL, avatarURL, description, url.

Default daily-test prefill

The actor's saved prefill is mode=topStreams, language=EN, maxItems=10. This always returns ≥1 record because Twitch is always live somewhere — a strong daily-test default.

FAQ

Do I need a Twitch account or API key? No. The actor uses the public web Client-ID (the same one the Twitch site sends from incognito). Optional authToken input is supported for higher rate limits, but is never required.

Why don't I see subscriber counts? Subscriber counts and a few other fields are gated by Twitch behind a logged-in OAuth token. Provide an authToken (the value of your auth-token cookie on twitch.tv) to unlock them, but they are not returned in default unauthenticated mode.

Why does the actor sometimes engage Apify proxy automatically? Twitch's GraphQL endpoint occasionally rate-limits aggressive datacenter IPs. If the first attempt returns 0 records (e.g., 429 / 403), the actor lazily acquires an Apify proxy session and retries. Setting useApifyProxy=true up-front avoids the warmup hit.

Can I scrape chat history? No. Chat (IRC / EventSub) is a separate live-streaming surface and out of scope for this actor.

What's the difference between byCategory and topClips? byCategory returns currently-live streams in a category (use it to find people streaming a game right now). topClips returns top clips in that category over a chosen time window.

How do I find a category slug? Open https://www.twitch.tv/directory/category/<slug> — for example just-chatting, league-of-legends, valorant. Or use the actor's topGames mode and read the slug field.

Limitations

• Twitch's GraphQL endpoint may rotate persisted-query hashes; this actor side-steps that by sending raw GraphQL queries (no persistedQuery extension), so hash rotation can never break it. The schema itself is stable.
• Subscriber counts, paid emote slots, and other partner-only fields require an authToken.
• Per-clip exact rendering URLs (.mp4 files) require a gqlClipsAccessToken separate flow that's beyond the scope of this actor; we surface the official embed URL instead.

Pricing

Pricing is configured on the Apify platform. There is no charging logic in the actor code itself.