YouTube Data API Alternative: Channels, Comments and Shorts via API and MCP

A YouTube Data API alternative is a way to read public YouTube information, such as a channel's videos, the comments under a video, a channel's shorts and the results for a keyword, without building on Google's own API. NeuralVerge's four YouTube sources do this with one request per question and return JSON with views, likes, durations, subtitles and subscriber counts. Reach them over REST or as MCP tools for AI agents.

The short version

• —Verdict. Use the official YouTube Data API when you need to act on a channel you own or manage. Use these sources when the job is reading public videos, comments, shorts and search results for research, at volume, with no Google Cloud project.
• —Who it is for. Marketing, product and sales teams doing category research, competitor tracking, review and objection mining and creator vetting, and developers building agents that need a YouTube step.
• —What it costs. Search and channel results are 5 points each ($5 per 1,000), comments and shorts are 2 points each ($2 per 1,000). A search that matches nothing costs nothing. See the pricing page for plans.
• —Where to look up exact fields. The YouTube search source page is the reference for every parameter, field and price. This post is about the workflow around it. The wider set of platforms is covered in the social media data guide.

When the official API fits, and when it does not

The official API: built for apps that act on YouTube

Google's API lets an application work with YouTube on a user's behalf: manage a channel's videos and playlists, read the data a channel owner is entitled to see, and embed YouTube features in a product. It runs on a Google Cloud project, an API key or OAuth credentials, and a quota system that limits how much a project can do per day. How the quota is counted and what each call costs are set by Google and change over time, so look them up in Google's current documentation; this article quotes none of them.

Where research workloads get friction

Teams who only want to read public information often describe the same friction: a cloud project and credentials before the first call, a second call for the statistics of each search result, quota to track across many channels or terms, and paginated comments. Check each against Google's documentation, because the details are theirs to define.

None of this is a defect. The question is whether a read-only research task across many channels wants a different shape of tool. If you need private analytics or want to upload and manage content, it does not.

The research workflow

NeuralVerge exposes YouTube through four data sources. Each takes one kind of input and returns one kind of record.

The source pages state the rates as $5 per 1,000 for search and channel results and $2 per 1,000 for comments and shorts, at the time of writing.

1. Choose the source that matches what you have

A keyword means YouTube search. A channel handle means YouTube channel or YouTube shorts. A video URL means YouTube comments, or YouTube channel for that single video with its subtitles.

2. Send one request with your limits

You send a JSON body to the source's endpoint with an API key. The limits you set decide how much you get back and spend: how many videos, shorts, streams or comments. Search filters (upload date, type, length, features such as subtitles) and sorting by relevance, rating, date or views narrow the result before you pay for it.

3. The lookup runs inside the request

Every data source endpoint is synchronous. The call blocks until the lookup finishes and returns the records in the same response, with no job to poll. If a lookup reaches its time budget first, the response comes back with partial: true and whatever was collected so far. Re-run with smaller limits to get a complete result.

4. Read records stamped with where they came from

Each record carries the URL it was read from, and each response a session id you can trace.

5. Billing follows what was returned

You are billed in points per record actually returned, capped by your limits. The source pages state that a search that matches nothing costs nothing and that a video with comments turned off costs nothing. Check each source page for its current billing note.

One real request and response

The featured source is YouTube search. This request looks for videos about CRM software reviews uploaded in the past year, between 4 and 20 minutes long, sorted by views.

curl -X POST https://api.neuralverge.ai/functions/v1/run-youtube-search \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "query": "crm software review", "max_results": 20, "date_filter": "year", "length_filter": "between420", "sorting_order": "views" }'

The response below was captured from the same endpoint for the same query with max_results set to 1, so it holds one result and costs 5 points. Asking for 20 or 1,000 results gives the same structure with more objects in items.

{
  "session_id": "6803169b-e36b-47e8-bd1d-f3646797f9d1",
  "kind": "youtube_search",
  "query": "crm software review",
  "max_results": 1,
  "max_results_shorts": 0,
  "max_results_streams": 0,
  "download_subtitles": false,
  "partial": false,
  "machine": {
    "total_items": 1,
    "items": [
      {
        "id": "RIRyuq-AKOU",
        "url": "https://www.youtube.com/watch?v=RIRyuq-AKOU",
        "type": "video",
        "title": "Best CRM Software for Small Business for 2026 (Top 5 Tested & Compared)",
        "description": "This videos offers insights into the best crm for small business options, helping you navigate the world of customer relationship management with ease. …",
        "hashtags": [],
        "date": "2025-10-29T20:37:38.000Z",
        "duration": "00:11:21",
        "view_count": 50993,
        "likes": 537,
        "comments_count": 39,
        "comments_turned_off": false,
        "is_paid_content": false,
        "subtitles": [],
        "channel": {
          "id": "UCSPh1bFTdDalKnZPgQoOYUg",
          "name": "LYFE Accounting",
          "username": "LYFEAccounting",
          "url": "https://www.youtube.com/@LYFEAccounting",
          "subscribers": 596000
        },
        "order": 0,
        "found_on_url": "https://www.youtube.com/results?search_query=crm+software+review&sp=EgIQAQ%3D%3D"
      }
    ],
    "errors": []
  },
  "total_points": 5
}

Reading the response

• —The video record is complete. Title, description, publish date, duration, view count, likes and comment count come back in one object. There is no second call to fetch statistics for each result.
• —The channel is attached. Each result names the channel and gives its subscriber count. For a category search, this tells you which creators rank for the term and how large their audience is.
• —subtitles is empty here because the request did not switch on download_subtitles. It holds subtitle tracks when you do, and stays empty for videos without captions.

Two limits apply to search. Results follow YouTube's ranking for an anonymous visitor in one region. And channel details on search results are lighter than on a channel lookup: total videos and total views can be null. For the full channel record, run the channel source on that handle.

Comments under a video

YouTube comments takes a video URL and returns comments and replies. Each record has the author handle, the text, likes, reply count, whether the creator hearted it and whether the channel owner wrote it. The source page's captured response, from the same CRM review video, cost 4 points for two comments. One of them, with the author handle shortened in the captured data:

{
  "author": "@beth…",
  "text": "Felt like a review video at first, then became an obvious GHL advert.",
  "published_time_text": "6 months ago",
  "vote_count": 27,
  "reply_count": 1,
  "has_creator_heart": false,
  "author_is_channel_owner": false,
  "video_comments_count": 39
}

Publish times come back as relative text, not exact timestamps, and authors are handles only. Comments are written by private individuals, so use them for legitimate business purposes under your own lawful basis and honour deletion and opt-out requests.

A channel's videos, with transcripts

YouTube channel takes a handle, a playlist URL or a single video URL. It returns regular videos and, if you set those limits above zero, shorts and live streams too. Set download_subtitles and a format, and the text of what was said comes back with each video where captions exist. This turns webinars, demos and customer stories into text that can be searched and quoted. The channel record includes subscribers, total videos, total views, links and the joined date.

A channel's shorts

YouTube shorts returns only the short-form videos, at 2 points each instead of 5. It is the cheaper choice when the question is how a channel performs on shorts, for example comparing views per short across three competitors. It has no subtitles; for transcripts, use the channel source.

Worked example: mapping a category

Imagine Acme Oy, a Finnish software company, preparing to enter a new product category. This is an illustration, not a real result, and there are no figures in it. A researcher runs YouTube search on four category terms and keeps the channels that appear for more than one term. For each of those channels, YouTube shorts shows whether short-form is working for them. For the three most-watched review videos, YouTube comments shows what viewers pushed back on. In an afternoon, Acme Oy has a map of who shapes opinion in the category and what viewers doubt.

REST and MCP

The search request above is also available as the MCP tool run_youtube_search. The other three are run_youtube_channel, run_youtube_comments and run_youtube_shorts. Tool names use underscores; REST paths use hyphens. The arguments mirror the REST body:

{ "tool": "run_youtube_search", "arguments": { "query": "crm software review", "max_results": 20 } }

Over REST you call the endpoint from a script or scheduled job. Over MCP, an agent calls the tool mid-task: search a category, choose the best video, read its comments, summarise with the source URLs attached. Same data, same points either way. For setup in Claude and other clients, see the MCP server guide. The reference for endpoints and parameters is at https://docs.neuralverge.ai, and the data sources page shows the current MCP status of each source.

Cost and limits

The NeuralVerge side

Every figure here comes from the source pages. Search and channel results are 5 points each, comments and shorts 2 points each. Totals are capped by the limits you set, and the source pages say a request that returns nothing is not billed. There is no Google Cloud project and no quota to monitor; what you manage is your point balance. Confirm current plans on the pricing page.

The official API side

We do not restate Google's quota or the cost of any call: those figures are Google's, they change, and a wrong number does real damage when someone sizes a project from it. Look them up in Google's documentation and do the same sum for both sides with your real volume.

Limits and pitfalls

• —Public data only. Watch time, retention, traffic sources and audience demographics belong to the channel owner. These sources do not return them, and the official API will not either unless you are authorised on that channel.
• —A snapshot at read time. Counts are as they were at the request. For trends, run it on a schedule and store the results.
• —Rounded likes. Like counts can be rounded the way YouTube displays them. The channel source's sample shows a like count of exactly 10000 on a video with hundreds of thousands of views.
• —Captions only where they exist. Subtitles come back only for videos that have them.
• —Terms of service. Whether a particular use is permitted depends on the platform's current terms, your jurisdiction and what you do with the data. This is not legal advice, so have counsel check the specifics.

How to decide

Whether you pick the official API, NeuralVerge or something else, answer these with your own test before you commit.

1. —Is the job reading or acting? If you need to modify a channel or read private analytics, you need the official API with proper authorisation.
2. —What volume do you need per day or week? Work out how many searches, channels and comments that implies, then price it on each option's current published terms.
3. —Do you need transcripts? Check how each option handles captions for videos you do not own, and what happens when a video has none.
4. —Do you need stats with the search results? Count the calls needed for views, likes and channel size per result.
5. —How will it be called? A script, a scheduled job and an agent tool loop have different needs. Check whether an MCP tool exists and its current status.
6. —How does it handle failure? Look for partial results and empty results, and how each is billed.
7. —What are the compliance requirements for comments? Decide your lawful basis, retention and deletion process before you store comment text.
8. —Does your use case match one of these?
   • —Category video landscape: which channels rank and how much the videos get watched.
   • —Competitor video strategy: a competitor's most popular topics, with subtitles for key videos.
   • —Review and objection mining: comments under reviews, grouped into objections, praise and questions.
   • —Creator vetting before a sponsorship: subscribers, upload cadence, views per video and per short.
   • —Agent research steps: search, pick, read comments, summarise, with sources.
9. —Run the same test on both. Take ten real channels and five real keywords and compare results, cost and integration time.

Frequently asked questions

Is this a replacement for the official YouTube Data API?

Not for everything. The official API is the right tool when you need to act on a channel you own or are authorised to manage, such as uploading videos or reading private analytics. The NeuralVerge YouTube sources cover the other job: reading public videos, comments, shorts and search results as structured JSON for research, with no Google Cloud project to set up.

What does a YouTube search request cost?

YouTube search is 5 points per returned video, short or stream, which the source page lists as $5 per 1,000 results. The total is capped by the limits you set in the request, and a search that matches nothing costs nothing.

Can I get comments and transcripts, not just video titles?

Yes. YouTube comments returns comments and replies under a video at 2 points each. YouTube channel and YouTube search can return subtitles when you switch on download_subtitles, but only for videos that have captions. Videos without captions come back with an empty list.

Is the YouTube data available over MCP?

Yes. The four YouTube sources are available over REST and as MCP tools: run_youtube_search, run_youtube_channel, run_youtube_comments and run_youtube_shorts. An agent can call them in a loop, for example search, pick a video, then read its comments. The data sources page shows the current MCP status for each source.

How do the quota and pricing compare with the official API?

They are different models. Google's API works on a quota system, and the exact numbers and rules change, so check Google's current documentation rather than relying on a figure in a blog post. NeuralVerge bills per returned result in points, with the rates shown on each source page and on the pricing page.

What YouTube data can I not get?

Anything private to the channel owner, such as watch time, retention and audience demographics. The sources return public information only. Comment publish times come back as relative text such as 9 months ago, and search results follow YouTube's ranking for an anonymous visitor in one region.

Is it allowed to collect public YouTube comments and video data?

It depends on the platform's current terms, your jurisdiction and what you do with the data, so this is not legal advice and your counsel should check your use. The general posture on the source pages is to read public information only, use it for legitimate purposes under your own lawful basis, and honour deletion and opt-out requests.