A YouTube scraper API returns YouTube search results as structured rows without the Data API — video id, title, channel and channel URL, view count as shown and as a number, publish age, duration, description snippet, badges and thumbnail.
The reason this exists is a single number: 100. A search call against YouTube's official Data API costs 100 quota units, and a default project gets 10,000 units a day. That is one hundred searches per day, total, before anything else your application does. Teams doing continuous topic monitoring hit that wall in the first afternoon and then spend a week writing a quota-increase justification. This reads the results page's own data payload instead — no key, no quota, no browser.
$0.0008 per delivered video, up to 40 videos per run. Nothing delivered means nothing charged.
A query goes to YouTube from an exit in your country and the results page's embedded data payload is parsed into rows. One results page per run, which is roughly 20 videos — YouTube puts everything past the first page behind a continuation token, and honest scoping is better than pretending otherwise.
sort is passed through to YouTube, so you can ask for relevance, upload date, view count or rating and get YouTube's own ordering rather than re-sorting a relevance list locally and calling it a date sort.
Views arrive twice on purpose: views as the display string YouTube rendered and views_value as an integer. The string is what you show a human; the integer is what you do arithmetic on. badges carries the labels YouTube attaches — new, 4K, CC and similar — which are useful filters and a nuisance to parse from HTML.
query is required. sort maps to YouTube's own ordering options rather than to a client-side re-sort. country and lang set the exit and the interface locale, which changes results more than people expect — YouTube's ranking is heavily regionalised. max_results caps delivery within the single page a run covers.
One row per video. video_id is the canonical id you would use anywhere else, and url the ready-made link. views and views_value are the display string and the integer. published is YouTube's relative age ("3 weeks ago") rather than a timestamp, because that is what the search page publishes — treat it as approximate. live flags live streams, which otherwise pollute view-count analysis badly, since a live stream's counter means something different. duration is null for live content for the same reason.
$0.0008 per delivered video ($0.80 per 1,000). Nothing delivered means nothing charged, and the $2 monthly free credit covers roughly 2,500 videos here. Volume tiers take up to 30% off.
$0.0008 per delivered video, 80 cents per 1,000. Monitoring 300 topic queries daily at 20 videos each is 6,000 rows, about $4.80 a day — and there is no quota to apply for, so scaling that to 3,000 queries is a budget decision rather than a support ticket.
Blocked pages and retries are never billed. The free $2 monthly allowance covers about 2,500 videos.
Be clear about when the official API wins: if you are managing your own channel, reading your own analytics, uploading, or fetching metadata for known video ids at modest volume, the Data API is free, well documented and correct. Use it.
The friction is search. A search.list call costs 100 quota units against a default daily allowance of 10,000 — about a hundred searches a day for the whole project. Simple videos.list reads cost 1 unit, which is why quota rarely bites until you start searching. Quota increases are available by application, and the process is neither instant nor guaranteed.
There is also a subtler gap. The Data API returns YouTube's API ranking for a query; the results page returns what a viewer in that country actually sees, including badges and ordering that the API does not model. For rank tracking and competitive monitoring, that difference is the whole measurement.
YouTube renders its results from an embedded JSON payload, so scraping it is more tractable than it looks — the problems are elsewhere:
For a channel's uploads rather than a search, the YouTube channel collector is the right endpoint.
Track which videos hold the top positions for your target queries per country, and whether yours appear at all — using the ordering viewers actually see.
Watch what competitors publish against your keywords and how it performs, without a quota ceiling on how many keywords you can watch.
Sort by view count to find what has worked in a topic, or by upload date to see what the space is publishing right now.
Current video results with channels, dates and view counts give a model something concrete to cite instead of a training-set memory.
One results page per run, roughly 20 videos, up to 40. That is a real limit and worth designing around: broader coverage comes from more queries rather than deeper pagination, since everything past page one sits behind a continuation token. More queries parallelise up to your plan's rate limit — 60 requests/minute on pay-as-you-go, up to 1,200 on the top tier.
Legally: search results and video metadata are publicly displayed, and collecting public data is generally lawful in most jurisdictions, but YouTube's terms restrict automated access. This returns metadata and links, never media — video content is copyrighted by its creators. Channel names can be personal data. Not legal advice.
No. Your QuantumProxies key is the only credential. There is no Google Cloud project, no OAuth and no quota to manage — which is the point, since a single Data API search costs 100 of a default 10,000 daily quota units.
There is a free allowance of $2 a month with no card, about 2,500 delivered videos at $0.0008 each. The YouTube Data API is also free, but its search quota works out to roughly 100 searches per day per project, which is the constraint most people run into.
$0.0008 per delivered video — 80 cents per 1,000 — with no subscription and no quota application. Volume tiers take up to 30% off, and blocked pages are never billed.
One results page, roughly 20 videos and up to 40. Everything past the first page sits behind YouTube's continuation token, so wider coverage comes from more queries rather than deeper paging.
You get both: views as the display string YouTube rendered, and views_value as an integer parsed from the payload. Use the string for display and the integer for arithmetic.
No. This collector covers search results. Comments and transcripts are separate surfaces with separate cost and copyright considerations.
Use the YouTube channel collector for that — it takes a channel handle or id and returns its uploads, shorts or streams tab, stamping channel-level stats on every row.
Skip the marketing. QuantumProxies.io publishes a machine-readable site map, Markdown for every page, a free MCP server, and APIs billed from the same balance as your proxies.
Connect in one command: npx -y quantumproxies-mcp