Loading ToolBench
An MCP server that provides tools for searching and fetching metadata from various external services including TMDB, ThePornDB, Metatube, Wikipedia, and DuckDuckGo.
This server has serious definition quality gaps across all tools. While tool names start with action verbs (web_search, fetch, search_japanese_porn, etc.), the definitions lack sufficient depth for production use. Most tool descriptions are 50-100 characters (below the 194-char baseline), and critically, NO tool has an explicit input schema visible in the source code provided. The tools use a custom registration pattern via mcptools.AddTools() without exposing JSON Schema definitions. Parameter descriptions exist but are minimal (1-2 lines). Error handling is absent, no guidance on what to do if a tool fails. The server includes content-related tools (porn searches) that raise domain concerns, though they are read-only.
Fetches content from a specified URL. Can optionally convert HTML content to Markdown.
Finds content on TMDB by IMDB ID using external source lookup.
Searches for Japanese and Chinese pornographic content on Metatube using a given ID (番号), e.g., 'SSIS-698'.
Searches for movies on The Movie Database (TMDB) by name (required) and optional release year.
Searches for non-Japanese pornographic movies and scenes on ThePornDB.
Searches for TV shows on The Movie Database (TMDB) by name.
No explicit JSON Schema input definitions visible in source code. Tools are registered via mcp.AddTool() with minimal schema info. Cannot verify parameter types, required fields, or constraints. This violates the pattern:constrained-input and pattern:tool-description requirements.
Tool descriptions are too short (40-50 characters vs. 194-char baseline). Examples: 'Searches for Japanese and Chinese pornographic content on Metatube using a given ID' (80 chars) and 'Searches for TV shows on The Movie Database (TMDB) by name' (58 chars). Descriptions lack WHEN to use, WHAT happens on error, and prerequisite context.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 54 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Performs a web search using DuckDuckGo and returns the search results.
Retrieves the full content of a Wikipedia page given its exact title.
Searches Wikipedia for pages matching a given query and returns a summary of each result.
No error handling documentation. Tools do not guide LLMs on what to do if a search returns no results, if an API is unavailable, or if a network timeout occurs. No recovery hints or error classification.
No output schema documentation. LLMs cannot determine what fields are returned, their types, or how to chain results to other tools. For example, search_movies likely returns movie objects with IDs that could be used downstream, but this is not documented.
Parameter descriptions are minimal (1-2 words). E.g., 'the search query string' for web_search.query lacks examples of valid formats, length constraints, or what happens if empty. 'the id (番号) of the jav to search for' for search_japanese_porn.jav_id lacks format validation hints.
API credentials (TMDB_API_KEY, ThePornDB_API_TOKEN, METATUBE_API_KEY) are managed server-side via environment variables, which is correct. However, no documentation about rate limits, quota exhaustion, or graceful degradation if APIs fail.
Domain concern: Two tools explicitly search for pornographic content (search_japanese_porn, search_porn). While read-only and intended for adult content discovery, this may violate acceptable-use policies for hosted MCP services and could expose platform operators to legal/compliance risks. Consider adding age-gate warnings or moving to a separate, opt-in server.