Tubask

Tools overview

Tubask exposes exactly three MCP tools. Your assistant should pick one entry point per user goal — not chain search → metadata → transcript → summarize.

Decision flow

User message
    │
    ├─ Paste link / @handle / playlist / search text?
    │       └─► youtube_query (intent=auto or goal shortcut)
    │
    ├─ Understand ONE video (summary, chapters, quotes)?
    │       └─► summarize_video
    │
    └─ Need EXACT words (keyword, timestamp, time range)?
            └─► get_transcript

The three tools

youtube_query
Primary router — search, channels, playlists, comments, multi-video creator advice.
  • intent=auto + ref for pasted URLs
  • goal=advice for creator synthesis (one call, up to 8 videos)
  • goal=latest for recent uploads
  • goal=comments for top comments on a video
summarize_video
Deep dive on a single video — don't call this in a loop for channel advice.
  • depth=brief for quick overview
  • depth=detailed for long talks
  • Returns chapters, quotes, key_points
get_transcript
Captions only — keyword search, pagination, time slices.
  • search_in_transcript for quotes
  • format=timestamped for citations
  • offset/limit for long videos

goal vs intent

goal is the user-friendly shortcut your assistant can pass to youtube_query. It maps internally to an intent. Prefer goal in natural prompts; use intent when you need precise control.

goalTypical use
autoDetect from pasted ref (default)
adviceWhat does this creator teach about X? (multi-video)
latestRecent uploads from a channel
researchTopic search across YouTube
commentsTop comments on a video
playlistVideos in a playlist
infoChannel or video metadata

next_steps in every response

Each tool response includes a next_steps[] array — suggested follow-up actions with ready-made parameters. Your assistant should read these instead of inventing another tool chain.

Anti-pattern

Don't call summarize_video eight times for creator advice. Use youtube_query(goal="advice") once.

Full reference

youtube_query
All parameters, intents, ref formats, and examples.
summarize_video
depth, languages, timestamps, and return shape.
get_transcript
Search, pagination, time ranges, and formats.