YouTube Channel Videos & Views Scraper — No Blocked Paths

Turn monitoring on and every later run returns only the videos it has not returned before, so you are never charged for the same video twice.

Give it a handle. Get that channel's videos with view count, length and publish time — read only through paths YouTube's robots.txt allows.

Unofficial. Not affiliated with or endorsed by YouTube or Google.

🔴 What makes this one different

YouTube's robots.txt disallows /feeds/videos.xml and /youtubei/ — the channel RSS feed and the internal player API. Those two are what most YouTube scrapers are built on.

This Actor uses neither. It reads /@handle/videos and /watch?v=, which are not disallowed, and the URL builder physically refuses to construct a disallowed address — a test fails if it ever does.

The allowed path is also the better one: the videos tab returned 30 videos when measured on 2026-09-06; the RSS feed returns 15.

🔴 Every row tells you whose data it is

A YouTube channel page is full of recommendations from other channels. Taking the first "channelId" on the page gives you someone else's channel — a mistake that is invisible in the output, because the numbers still look like numbers.

So this Actor:

1. reads the channel id only from the page's own <link rel="canonical">,
2. reads video ids only from inside the videos grid, not the whole page,
3. and then opens each video and records the channel that video itself says published it, next to the one you asked for.

That last one is the column ownerMatchesRequested. If it is ever false, the row is telling you the data does not belong to the channel you named. When the owner cannot be read it is null, not false — unknown is not the same as wrong.

What you get in every row

Field	What it is
channelInput, channelId, channelTitle	What you asked for, and what it resolved to
videoId, videoUrl, title	The video
viewCount, lengthSeconds	Views and runtime
publishedAt, publishedAtRaw	ISO instant and the site's own string (2026-09-05T14:18:49-07:00), so you can check the conversion
ownerChannelId, ownerChannelName, ownerMatchesRequested	Who actually published it, and whether that matches
playabilityStatus, playabilityReason	Where the page reports one
videoKey	What monitoring compares

Pricing: per row, plus a monthly fee only if you monitor

• Per video returned — you pay for the rows you actually receive.
• $5 once per calendar month (UTC), charged on the first monitoring run of that month.

The monthly fee is what pays for watching on the days when nothing was posted. Those runs return no rows and cost you nothing else, so a channel that goes quiet for three weeks does not quietly stop being watched to save money.

Runs with monitoring turned off never trigger it — one-off pulls are per-row only. If a charge cannot be made (for example the run hits its own maximum total charge), nothing is recorded and it is attempted again on the next run, so it is never taken twice.

Monitoring: two meanings of "new"

• New videos only — a video is returned once, ever. Already-returned videos are skipped without opening their page, so a scheduled run on a quiet channel costs almost nothing.
• Any change — a video comes back when its view count or title changes, which is how you build a view-growth history.

Configure the run

{ "channels": ["@mkbhd"], "maxVideosPerChannel": 30, "maxVideos": 100 }

Follow several channels for new uploads:

{ "channels": ["@mkbhd", "@veritasium", "UCLA_DiR1FfKNvjuUpBHmylQ"], "monitoringMode": true, "monitorMode": "new-videos" }

Long-form only, from this year, tracking view growth:

{ "channels": ["@mkbhd"], "minLengthSeconds": 62, "publishedAfter": "2026-01-01", "monitoringMode": true, "monitorMode": "any-change" }

What this Actor refuses to guess

A mistyped handle comes back as no-such-channel, not a guess — YouTube's search (/results) is disallowed by robots.txt, so this Actor does not have a way to "find something close" and does not pretend to.

A bot check is reported as blocked, never as an empty result. Channels with no uploads, filters that matched nothing and unreadable pages each come back as their own row with the reason, and none of them are charged.

Input reference

Every field is optional; the defaults read the latest videos from one channel.

Field	Type	Default	What it does
channels	array	["@mkbhd"]	Handles (@mkbhd), channel ids (UC...) or channel URLs. A mistyped handle is reported as 'no such channel' rather than guessed, because YouTube's robots.txt disallows…
maxVideosPerChannel	integer	30	How far down the channel's videos tab to go. The tab holds about 30 videos (measured 2026-09-06); the RSS feed other scrapers use holds 15, and is disallowed by…
maxVideos	integer	100	How many rows to return in total. You are charged for the rows you actually receive. Each video costs one extra page request for its view count.
monitoringMode	boolean	false	Off = return the videos on the tab. On = remember what was returned and, on later runs, return only what is new. With 'new videos only' below, already-returned videos…
monitorMode	string	new-videos	New videos only = a video is returned once, ever. Any change = a video is returned again when its view count or title changes, which is how you build a view-growth…
resetMonitoringState	boolean	false	Clear the remembered videos before this run, so everything counts as new again.
minViews	integer	0	Keep a video only if it has at least this many views. 0 = no limit. Videos whose view count could not be read are kept, because unknown is not the same as low.
maxViews	integer	0	Keep a video only if it has at most this many views. 0 = no limit.
minLengthSeconds	integer	0	Keep a video only if it runs at least this long. 0 = no limit. Use 62 to skip Shorts.
maxLengthSeconds	integer	0	Keep a video only if it runs at most this long. 0 = no limit. Use 61 to keep only Shorts.
publishedAfter	string	(empty)	Keep only videos published on or after this date. Videos with an unreadable publish time are kept.
onlyOwnedByChannel	boolean	false	Every row already reports whether the video's own page names the channel you asked for. Turn this on to drop the ones that do not match. Rows where the owner could…
keywords	array	[]	Keep a video only if one (Any) or all (All) of these appear in the title or channel name.
keywordMatch	string	any	Any = at least one keyword. All = every keyword.
excludeKeywords	array	[]	Drop a video if any of these appear in the same fields.
useProxy	boolean	false	Off by default. If YouTube answers with a bot check, the Actor switches to a proxy session on its own and tries again; turning this on from the start avoids that…

Pricing

Pay per result: you are charged for the video rows you actually receive, plus the $5 monthly access fee described above only when monitoring mode is on. Reading pages is free; explanation rows are never charged.

Every number in this README is checked

The fixtures in test/fixtures/ are the real pages this Actor was built from (2026-09-06). The test suite re-counts every figure quoted above — 30, 15, 3 — and fails if a disallowed URL can ever be built.