youtube-research

YouTube research

Resolve YouTube videos and channels, and pull transcripts, comments, playlists, and search results as normalized JSON from the Crawlora API — no yt-dlp, no browser, no scraping.

When to use this skill

• The user pastes a YouTube URL and wants its transcript, summary, or stats.
• "Get the comments on this video" / "what are people saying?"
• "List this channel's recent videos / shorts / playlists."
• "Search YouTube for …" or "what videos exist about …".
• Building a transcript-based summary, sentiment, or research pipeline.
• Comparing a bounded set of videos with timestamped transcript evidence.

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 id. A YouTube watch URL https://www.youtube.com/watch?v=<ID> (or https://youtu.be/<ID>) gives the video id; channel URLs give a channel id. Extract the id before calling.
2. Transcript: /youtube/transcript/{id} (use /youtube/transcript/{id}/languages first if you need a specific language).
3. Metadata / comments: /youtube/video/{id}, /youtube/comments/{id}.
4. Channel: /youtube/channel/{id}/videos, /channel/{id}/shorts, /channel/{id}/playlists, /youtube/profile/{id}.
5. Discovery: /youtube/search?q=... to find videos by keyword.

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

Transcript-backed topic research

For a topic brief, search first, deduplicate by video ID, and choose a bounded sample that reflects the requested date, language, and video types. Discover available transcript languages before fetching captions. Request JSON with timestamps for evidence extraction; tie each theme or claim to the actual caption segment and video URL. A caption or translation may mishear speech and does not independently verify factual claims. Report unavailable transcripts and keep search rank, view counts, and comment samples separate from demand.

For a wider map of topic and format coverage across videos or channels, use youtube-content-gap-analysis when installed.

scripts/crawlora.sh /youtube/search q="home energy storage" type=video sort_by=relevance
scripts/crawlora.sh /youtube/transcript/dQw4w9WgXcQ/languages
scripts/crawlora.sh /youtube/transcript/dQw4w9WgXcQ format=json timestamps=true

Calling the API

# Transcript for a video id (GET):
scripts/crawlora.sh /youtube/transcript/dQw4w9WgXcQ | jq '.'

# Video metadata + comments:
scripts/crawlora.sh /youtube/video/dQw4w9WgXcQ | jq '.data'
scripts/crawlora.sh /youtube/comments/dQw4w9WgXcQ | jq '.data'

# Search:
scripts/crawlora.sh /youtube/search q="web scraping tutorial" | 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 YouTube endpoint this skill uses (method, path, params, description).

Examples

• Summarize a video: extract the id from the URL → /youtube/transcript/{id} → concatenate transcript segments → summarize.
• Comment sentiment: /youtube/comments/{id} → bucket comments by sentiment and surface common themes.
• Channel digest: /youtube/channel/{id}/videos → list recent uploads with view counts and publish dates.

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 YouTube pages; respect YouTube's terms.
• Security: key lives in CRAWLORA_API_KEY only — never hardcode, query-param, or commit it.
• Not every video has captions; check /youtube/transcript/{id}/languages when a transcript request returns empty.