YouTube MCP server for searching videos and getting transcripts
Two tools with basic verb-noun naming (search_youtube_videos, get_youtube_transcript). Both have descriptions (60-80 chars, below the 194-char baseline). Input schemas are present with type definitions and parameter descriptions. However, output schemas are not documented, tools return unstructured dicts with error fields mixed into success responses. No error classification or recovery guidance. Missing tool annotations (readOnlyHint). Parameter descriptions are minimal (10-30 chars vs 72-char baseline). No pagination support despite search returning variable results. Error handling returns raw exception strings rather than actionable guidance.
Get the transcript of a YouTube video.
Search YouTube videos based on a search term.
Output schemas not documented. Tools return dicts with mixed error/success fields (e.g., {"error": "..."}). LLMs cannot plan downstream calls or extract structured data reliably.
Error responses return raw exception strings ('Failed to search YouTube videos: ...') without recovery guidance. LLM cannot determine if error is retryable, user-fixable, or fatal.
Parameter descriptions are minimal (10-30 chars vs 72-char baseline). 'The search query for YouTube videos' and 'YouTube video URL or video ID' lack format constraints, examples of valid input, or dependency hints.
search_youtube_videos lacks pagination. Returns up to 50 results with no limit parameter, offset, or next_cursor. Large result sets blow context window; no guidance on result count.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 46 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Both tools are read-only but this is not declared in the schema, forcing LLMs to infer safety from descriptions.