YouTube Search → Video IDs

Search terms in, a flat list of YouTube videos out: video ID, URL, title, channel, views, duration and upload age. One row per video.

Which YouTube Actor should I use?

This is a narrow Discover step. It finds videos for a keyword and stops there. It does not download transcripts, comments, likes, full descriptions, Shorts or channel pages. Hand the videoId/url it returns to the next step:

Next step	Actor
Captions as timestamped JSON	YouTube Transcript Extractor
Comments	YouTube Comments Extractor
RAG chunks from subtitles	Subtitle RAG Chunker
All videos of one channel	YouTube Channel Videos

Input	queries[], optional maxResultsPerQuery, sortBy, uploadDate, duration, language, country
Output per video	videoId, url, title, channelName, channelId, channelUrl, viewCount, durationSeconds, publishedTimeText, publishedAtEstimate, isLive, descriptionSnippet, thumbnailUrl, query, rank
Pricing	$0.00788 per video listed ($7.88 / 1,000). Queries with no results are not billed.
Access	YouTube's public web search endpoint. No login, no API key, no browser.
Call it	REST skilled_glee~youtube-search-videos · MCP https://mcp.apify.com/?tools=skilled_glee/youtube-search-videos

Pipeline role (n8n / agents): Content → Discover. Search → Transcript → Chunk is one n8n flow. Full catalog: EchoEmpire Machine Foundry llms.txt.

Input

{
  "queries": ["n8n tutorial"],
  "maxResultsPerQuery": 5,
  "sortBy": "relevance",
  "uploadDate": "month",
  "duration": "medium"
}

• queries (required, 1–50).
• maxResultsPerQuery (1–500, default 50). Results come in pages of about 20; the Actor stops at your limit or when YouTube has no more results.
• sortBy: relevance (default) or view_count. YouTube no longer honors an "upload date" sort, so it is not offered; use uploadDate to get recent videos.
• uploadDate: any, today, week, month, year.
• duration: any, short (< 4 min), medium (4–20 min), long (> 20 min).
• language (hl, default en) and country (gl, default US).

Output

Row from a real run (2026-10-29):

{"query": "n8n tutorial", "rank": 1, "videoId": "xl7Y49yLPDs", "url": "https://www.youtube.com/watch?v=xl7Y49yLPDs", "title": "n8n Tutorial for Beginners in 2026: How to Build AI Agents", "channelName": "Youri van Hofwegen", "channelId": "UCdR7eGozLjafYMTd2BFnJaQ", "channelUrl": "https://www.youtube.com/@Yourivanhofwegen", "viewCount": 44366, "viewCountText": "44,366 views", "publishedTimeText": "2 weeks ago", "publishedAtEstimate": "2026-10-16", "durationSeconds": 1052, "durationText": "17:32", "isLive": false, "descriptionSnippet": "Host n8n with Hostinger ...", "thumbnailUrl": "https://i.ytimg.com/vi/xl7Y49yLPDs/hq720.jpg", "status": "ok"}

Field notes:

• YouTube search shows upload time only as relative text ("2 weeks ago"). publishedTimeText is that text; publishedAtEstimate is computed from it at run time, so it is approximate (a date for day/week/month/year units). For exact dates, use a details/transcript step.
• viewCount is null for live streams (YouTube shows "watching" instead of views).
• rank is the position within that query's results in this run.

A query that returns nothing (or that YouTube refuses) produces one row {"query": "...", "status": "error", "error": "..."} and is not billed. SUMMARY in the default key-value store has per-query counts and whether a spending limit stopped the run.

Pricing (pay per event)

Event	When	Price
video-listed	Once per video row with status: "ok" saved to the dataset	$0.00788

No start fee. Error rows are free. Example: 4 queries × 50 videos = 200 × $0.00788 = $1.58.

Limits

• Regular videos only (the search runs with YouTube's "Videos" type filter); Shorts shelves, playlists and channels are skipped.
• Results are what YouTube's web search returns for an anonymous visitor from the Apify cloud for your language/country; they can differ from a signed-in search.
• YouTube can rate-limit requests; affected queries return error rows.