Loading ToolBench

Reading the current corpus…

YoutubeTranscriptMCP MCP server · B · ToolBench

← Browse

YoutubeTranscriptMCP

github.com/n0an/YoutubeTranscriptMCP

Extract video IDs and transcripts from YouTube videos, and search the web.

2026-07-28+STREAMABLE_HTTP3 tools

77/100

Definition Quality76

Protocol Readiness78

SupportabilityNot measured-

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence · evidence: Streamable HTTP

Current-spec patterns detected

Summary

Server provides 3 tools with visible definitions, proper input schemas, and descriptions. However, there are notable gaps in output schema documentation, error handling guidance, and parameter constraints. Tools follow verb_noun naming but descriptions lack LLM-optimized structure. Parameters are typed but missing some validation constraints and range specifications. No tool annotations (readOnlyHint, idempotentHint) present despite all being read-only operations. Composition is reasonable, tools are single-purpose and chainable (video_id extractor feeds into transcript tool).

Tools (3)

get_youtube_transcriptread onlysource verified77/100

Get the full transcript of a YouTube video. Args: video_url_or_id: YouTube URL or video ID language: Language code for transcript (default: "en") Returns: Full transcript as formatted text

get_youtube_video_idread onlysource verified75/100

Extract YouTube video ID from various YouTube URL formats. Supports: - https://www.youtube.com/watch?v=VIDEO_ID - https://youtu.be/VIDEO_ID - https://youtube.com/watch?v=VIDEO_ID - And other common formats

search_webread onlyauthsource verified75/100

Search the web using Tavily API to get comprehensive information about any topic. Args: query: The search query string max_results: Maximum number of results to return (default: 5, max: 20) include_domains: Optional list of domains to include in search (e.g., ["wikipedia.org", "github.com"]) exclude_domains: Optional list of domains to exclude from search (e.g., ["example.com"]) Returns: Formatted search results with titles, URLs, and content snippets

Issues (6)

HIGHtool ↗

Output schemas not formally documented for any tool. LLMs cannot plan downstream calls or structure parsing logic without knowing return types and field names.

get_youtube_video_idget_youtube_transcriptsearch_web

MEDIUMtool-description ↗

Tool descriptions lack LLM-optimized structure. Docstring format (Args/Returns) is readable to humans but wastes tokens and buries selection context. Should be 10 - 200 character narrative with WHAT, WHEN, and key output fields.

get_youtube_transcriptsearch_web

MEDIUMtool-annotations ↗

No tool annotations (readOnlyHint, idempotentHint, destructiveHint) despite all three tools being read-only and idempotent. This metadata aids agent planning and safety heuristics.

get_youtube_video_idget_youtube_transcriptsearch_web

MEDIUMconstrained-input ↗

Recommendations

Add formal output schema documentation to each tool. E.g. 'Returns: object {video_id: string (11 chars, alphanumeric and underscore/dash only)}'
Rewrite descriptions as LLM-optimized narratives: '[Tool action]. Use this when [context]. Returns [key fields].' Target 50 - 200 characters. Example: 'Extract the 11-character YouTube video ID from a URL or raw ID. Use before fetching transcripts. Returns the ID as a plain string.'
Add tool annotations to all three tools: { readOnlyHint: true, idempotentHint: true } in FastMCP tool registration.
Add JSON Schema constraints to numeric/enum parameters: max_results should include 'maximum': 20 and 'minimum': 1 in the schema. Language should be documented as ISO 639-1 code (e.g., 'en', 'es', 'fr') or constrained to an enum of supported languages.
Restructure error messages to be actionable and classified. E.g., for transcript not found: 'Retryable: Transcript not found for language "en". Call this tool again with language from the available list: es, fr, de.' For API key missing: 'Fatal: search_web requires TAVILY_API_KEY environment variable. Set it before starting the server.'
Document TAVILY_API_KEY requirement in the search_web description: 'Requires TAVILY_API_KEY environment variable. Set before server start.'
Consider adding a discovery/help tool (e.g., 'get_available_transcript_languages') to let agents query which languages are available for a video without trial-and-error.
Add page/limit pattern to search_web if Tavily API supports pagination. Currently max_results is capped at 20; if the LLM needs more, it cannot paginate.

Score history

Overall score trend

↑ 25 points across a rubric change (v1 → v2)

77/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	B	77	2026-07-28+	v2
2026-03-09	D	52	-	v1