Tubask

YouTube MCP server

YouTube research from chat

A YouTube MCP server exposes search, summaries, and captions to AI clients over the Model Context Protocol. Tubask hosts one at tubask.app/mcp with three tools and OAuth signup.

Last updated:

Hosted vs local →

3 summaries · 25 searches · about 5 minutes

tubask.app/mcp

Streamable HTTP · OAuth

  • youtube_query — search, channels, playlists
  • summarize_video — thesis, chapters, takeaways
  • get_transcript — quotes with timestamps

Three tools. Not twenty.

Research, summaries, and quotes — without six-call chains.

  • youtube_query

    Find videos, scan channels, and list playlists from a pasted link or search term.

  • summarize_video

    Get the thesis, chapters with timestamps, and takeaways from one link.

  • get_transcript

    Pull the exact words and timestamp from captions.

Next: How it works

How it works

What is a YouTube MCP server?

An MCP server that returns YouTube captions and metadata to your assistant. You paste a link in chat instead of copying transcripts by hand.

Get started

Claude, Cursor, ChatGPT, or Windsurf — same hosted endpoint.

  1. 01

    Create account

    Sign up at tubask.app. Trial includes 3 summaries and 25 searches.

  2. 02

    Add to your client

    Paste the MCP URL in Cursor, Claude, or ChatGPT settings.

  3. 03

    Paste a link

    Ask for a summary, channel scan, or quote with timestamp.

Paste into

.cursor/mcp.json

Project folder, or `~/.cursor/mcp.json` for all projects.

{
  "mcpServers": {
    "tubask": {
      "url": "https://tubask.app/mcp"
    }
  }
}
  1. 1.Copy the JSON below
  2. 2.Save the file, then Settings → Tools & MCP → Connect
Full setup guide →
Hosted vs local MCP →

Common questions

Start with one link.

3 summaries · 25 searches. No credit card — you'll know from the first reply.

About 5 minutes to connect Claude or Cursor