YouTube MCP server

How an agent would call it

Two steps: search returns refs, then data_fetch opens one of them. The arguments below are the arguments that were sent.

{
  "tool": "search",
  "arguments": {
    "query": "model context protocol tutorial",
    "sources": [
      "youtube"
    ],
    "limit": 2,
    "runId": "<from run_begin>"
  }
}

{
  "tool": "data_fetch",
  "arguments": {
    "sourceId": "youtube",
    "ref": {
      "id": "video:tzrwxLNHtRY"
    },
    "runId": "<from run_begin>"
  }
}

This runs against an account — send Authorization: Bearer <token> with the call. Get one at app.bedrocktap.com/dashboard/tokens; the free tier needs no card, and it reaches youtube like every other source.

What it costs. 1 credit per search call and 1 per fetch, whatever the row count. The search above returned 2 videos for 1 credit; the fetch cost 1.

What came back

{
  "ref": {
    "id": "video:tzrwxLNHtRY",
    "url": "https://youtube.com/watch?v=tzrwxLNHtRY",
    "title": "Model Context Protocol Clearly Explained | MCP Beyond the Hype"
  },
  "text": "Model Context Protocol Clearly Explained | MCP Beyond the Hype",
  "fetchedAt": "2026-09-12T14:58:45.297Z"
}

Each ref carried kind, channel, views, published and lengthText in its meta — 567,138 views over 15:04 for the first hit, 1,299,105 for the second. The fetched payload carried the video record: description, keywords, viewCount, likeCount, lengthSeconds, publishDate, category, channelId, channelTitle, hasCaption, subtitles, thumbnail and the availability flags.

Observed 2026-09-12. One call against a live source — ask again tomorrow and the rows move.

What a row contains

One row is one video, channel or playlist. The id is prefixed with its kind — video:abc123, channel:abc123, playlist:abc123 — because a fetch has to know which record to open. A video row carries the channel name, the view count, a relative publication date and the running time; a channel row carries subscriber and video counts; a playlist row carries its length and its channel. Set type on the call to choose which of the three you get.

A question it answers. Who has explained this on video, how long is it, and how many people watched?

Routing guidance

What an agent reads to decide whether youtube is the right source, and what to reach for when it is not.

Answers

Tutorials, product demos, long-form opinion and comment sentiment. The transcript path is the valuable one: search for videos, then fetch a result for the full transcript and comments — that turns video into a searchable text corpus.

Query shape

Plain words. Video, channel or playlist is configurable.

Price

Priced in credits, per row, per query or per call depending on the source. Every call is quoted before it runs — `cost_estimate` returns the exact figure and spends nothing. See /pricing.

Latency

Fast (~1-3s) for search; transcripts add a few seconds.

Prefer instead

`youtube-contacts` for channel contact details.

Cannot answer

Channel contact emails, view-count history, or anything behind age or region gates.

Related sources