Tubask
v0.5.0

Tubask documentation

Everything you need to connect YouTube to Claude, Cursor, or any MCP client — search videos, summarize talks, and pull exact quotes without leaving chat.

What is Tubask?

Tubask is a hosted MCP (Model Context Protocol) server for YouTube. Your AI assistant calls three tools — youtube_query, summarize_video, and get_transcript — instead of chaining a dozen granular API wrappers.

Paste a link, @handle, playlist URL, or plain search text. The server routes your request, returns structured results, and includes next_steps[] so your assistant knows what to do next — without guessing.

Who is this for?

  • Researchers & learners — turn long lectures into chapters and takeaways inside chat.
  • Creators & marketers — track channels, read comments, and synthesize what a creator teaches across videos.
  • Writers & journalists — find exact quotes with timestamps for verifiable citations.
  • Developers using Claude Code or Cursor — research tutorials and compare approaches without opening YouTube in a browser tab.

Three tools, not twenty

Most YouTube MCP servers expose 10–20 tools. Models chain them unpredictably, burning tokens and YouTube API quota. Tubask collapses the surface area:

youtube_query
Starting point for almost everything — links, handles, playlists, search text.
  • What's the latest from @Fireship?
  • Research videos about RAG evaluation
  • Top comments on this video
  • What does this creator teach about pricing? (goal=advice)
summarize_video
Understand one video deeply — summary, chapters, quotes.
  • Break down this 45-minute conference talk
  • Main ideas + 3 quotable lines
  • Chapter markers for a long tutorial
get_transcript
You need the exact words — keyword search, time slices, pagination.
  • Find every mention of "latency" with timestamps
  • Pull captions from minute 12 to 18
  • Paginate through a long transcript

Rule of thumb

Paste anything → youtube_query. One video deep dive → summarize_video. Exact quote or keyword → get_transcript.

What you need

Transcript fetches do not consume YouTube Data API quota. Only search, metadata, channel listings, and comments use your daily units. See Quota & billing.

Explore the docs

Quick start
Sign up, add your API key, connect your client, and run your first query in ~5 minutes.
Connect your client
Copy-paste configs for Claude, Cursor, Gemini CLI, Windsurf, and generic HTTP MCP.
Tools reference
Full parameter docs for all three MCP tools with intents, goals, and examples.
Example prompts
Ready-to-send chat prompts for learning, research, competitive intel, and citations.
Troubleshooting
Fix OAuth expiry, invalid API keys, quota errors, and client config mistakes.
Security
How we encrypt keys, handle OAuth, and what data we never store.