twitch-research

Twitch research

Resolve Twitch channels, games/categories, and teams, and pull live status, clips, VODs, VOD chat replay, schedules, and search results as normalized JSON from the Crawlora API — no Twitch developer app, no OAuth, no scraping.

When to use this skill

• "Is <channel> live right now, and what are they playing?"
• "Get the top clips from <channel>."
• "What did chat say during <VOD>?" / VOD chat replay lookups.
• "What's <channel>'s upcoming schedule?"
• "Who's live in <game/category> right now?" or top games site-wide.
• "Who's on <team>'s roster, and who's currently live?"
• "Search Twitch for a channel or category."

Setup (one-time)

• Get a free Crawlora API key (2,000 credits/mo, no card) at https://crawlora.net.
• Set CRAWLORA_API_KEY in the environment before running the helper.
• The helper reads CRAWLORA_API_KEY from the environment and sends requests to https://api.crawlora.net/api/v1. Missing/invalid key → 401.

How it works

1. Resolve the channel login. A Twitch URL https://twitch.tv/<login>

gives the channel login directly (it's already the lowercase URL slug) — no separate id-lookup step is needed.

1. Profile + live status: /twitch/channel login=<login>.
2. Content for that channel: /twitch/clips login=<login> (recent

clips), /twitch/videos login=<login> (VODs), /twitch/schedule channel=<login> (upcoming broadcasts).

1. VOD chat: once you have a VOD id (from /twitch/videos or a

twitch.tv/videos/{id} URL), page through its chat replay with /twitch/vod-comments video=<id>, advancing offset from the last comment's offset_seconds to keep paginating.

1. Team rosters: /twitch/team team=<slug> for a team's members and

who's currently live.

1. Discovery (no channel needed up front): `/twitch/search

query=<text> to find channels/games by name; /twitch/top-games for what's trending site-wide; /twitch/streams game=<slug>` for the top live streams under one game/category.

Full endpoint list, methods, and params: `reference/endpoints.md`.

Calling the API

# Channel profile + live status:
scripts/crawlora.sh /twitch/channel login=caedrel | jq '.'

# Clips and VODs for a channel:
scripts/crawlora.sh /twitch/clips login=caedrel limit=10 | jq '.data'
scripts/crawlora.sh /twitch/videos login=lck limit=10 | jq '.'

# Top games, then top streams under one:
scripts/crawlora.sh /twitch/top-games limit=10 | jq '.'
scripts/crawlora.sh /twitch/streams game=league-of-legends limit=10 | jq '.'

# VOD chat replay, paginating by offset_seconds:
scripts/crawlora.sh /twitch/vod-comments video=1234567890 offset=0 | jq '.'

Use scripts/crawlora.sh for all requests; it keeps the API key out of command-line arguments.

Endpoint reference

See `reference/endpoints.md` for every Twitch endpoint this skill uses (method, path, params, description).

Examples

• Channel digest: /twitch/channel (live status) + /twitch/clips +

/twitch/schedule for one login → a single "what's this streamer up to" summary.

• VOD moment-finder: /twitch/videos to find the VOD id →

/twitch/vod-comments starting at offset=0, paging forward by the last comment's offset_seconds → scan chat for reactions around a timestamp.

• Category browse: /twitch/top-games for the top categories →

/twitch/streams game=<slug> on the one of interest → surface the top live streamers under it by viewer count.

Notes & limits

• Credits / pay-on-success: billed only on 2xx; free tier 2,000 credits/mo.

Key at https://crawlora.net.

• Public data only — public Twitch pages/GraphQL surface; respect

Twitch's terms.

• Security: key lives in CRAWLORA_API_KEY only — never hardcode, query-param, or commit it.
• login/channel params are the lowercase URL slug (e.g. caedrel for

twitch.tv/caedrel), not a display name — lowercase before calling.

• /twitch/schedule returns an empty segments list, not an error, when a

channel has no schedule configured.

• /twitch/search is sourced from Twitch's search-typeahead surface — a

capped suggestion list (limit max 30), not a fully paginated results page.

• /twitch/vod-comments paginates by re-requesting with offset set to the

last returned comment's offset_seconds — there's no separate page token.