YouTube Latest Videos: No API Key, With View Counts

The latest videos from any YouTube channel or playlist: about 30 recent videos per channel with their view counts, upload age and length, as clean JSON. Turn on exactStats for the exact view count, publication date and like count of each video.

No Google Cloud project. No Data API key. No quota. No browser. No cookies.

Why that matters

Every other way to read YouTube programmatically costs you something:

• The YouTube Data API needs a Google Cloud project and gives you 10,000 quota units a day. A single channel listing costs ~100. It does not survive real use, and when it runs out your pipeline stops until midnight Pacific.
• A headless browser is slow, expensive per run, and breaks every time YouTube reshuffles its markup.

This reads the ordinary public pages: a channel's own Videos tab and a playlist's own page, the ones robots.txt leaves open.

What you paste

Any of these. Whatever is in your address bar will work:

@mkbhd
mkbhd
https://www.youtube.com/@veritasium
https://www.youtube.com/channel/UCBJycsmduvYEL83R_U4JriQ
https://www.youtube.com/c/mkbhd
https://www.youtube.com/user/marquesbrownlee
UCBJycsmduvYEL83R_U4JriQ
https://www.youtube.com/playlist?list=PLbpi6ZahtOH6Blw3RGYpWkSByi_T7Rygb

What you get back

One row per video, measured on @mkbhd: 30 rows.

field	notes
title	as published
url	the watch link
videoId	YouTube's own id
viewCount	parsed from what YouTube prints ("9.9M views" is 9,900,000)
viewCountText	exactly what YouTube printed
publishedTimeText	YouTube's own wording, e.g. "4 days ago"
durationText	e.g. "14:32"
thumbnailUrl	full-size image
channelName, channelId, channelUrl	who published it
requestedTarget, sourceType	what you asked for, so a row traces back

With exactStats on, each video's own page adds the exact figures:

field	notes
viewCount	exact, and viewCountIsExact is true
publishedAt	the publication date
likeCount	likes
lengthSeconds	length in seconds

YouTube shows individual video pages to datacenter addresses only behind a sign-in wall, so exactStats needs the RESIDENTIAL proxy group. The Actor says so in the log before it spends a request, and keeps the rounded figures rather than failing if a page cannot be read.

The limit, before you pay for it

One page per channel: about 30 recent videos. A playlist page holds more (19 and 100 on the two measured). There is no paging.

This reads a channel's recent output. It cannot read a back catalogue. If you need a full history, you need the Data API and its quota.

Filters

Filtered videos are never returned and never charged.

• titleKeywords / titleExclude: plain text, case-insensitive
• minViews: compared against YouTube's rounded figure; a video with no published count is excluded when this is set, because no count is not the same as zero
• publishedAfter: an ISO date; it needs the exact publication date, so it turns exactStats on for the run
• maxVideosPerChannel: lower it to keep a trial cheap
• maxResults: hard ceiling across the whole run, 1,000 by default

Good for

• Watching competitors' channels for new uploads and how they perform
• Tracking a topic across a set of creators
• Feeding a newsletter, dashboard or alerting pipeline
• Measuring which of your own videos are still gaining views

Pricing

Pay per video returned. Nothing else: not the run, not filtered videos, not a channel that turns out to be empty.

Use it from an AI agent

This Actor is a ready-made tool for Claude, ChatGPT, Cursor or any other MCP client, through Apify's MCP server. Add it as a connector at:

https://mcp.apify.com?tools=glitchbound/youtube-latest-videos-scraper

In Cursor or VS Code the same address goes in mcp.json:

{ "mcpServers": { "apify": { "url": "https://mcp.apify.com?tools=glitchbound/youtube-latest-videos-scraper" } } }

The first call asks you to sign in to Apify, or takes an API token. From then on the agent sees this Actor's input fields and calls it directly.

Feedback

If this saved you time, a review on the Actor's Store page helps other people find it. If something is missing or wrong, the Issues tab on the same page is the fastest way to reach me, and it gets read.