This Wikipedia server has severe definition quality gaps. All five tools lack parameter descriptions entirely (critical per pattern:tool-description). Tool descriptions are missing or trivial (all 0 chars, visible only as docstring placeholders in FastMCP decorator). No output schemas are documented. Parameter naming is inconsistent: 'query' appears in four tools but is never described; 'lang' in set_lang is also undescribed. Four tools (search, summary, page, random) return unstructured Wikipedia API responses without pagination, limits, or field documentation. The set_lang tool has a side effect (WRITE risk) but no error handling or confirmation pattern. No tool declares its output schema. This server requires fundamental documentation work before use in production.

• Add docstrings to all @mcp.tool() decorators. Each tool must have a 50 - 200 character description explaining WHAT it does, WHEN to use it, and any prerequisites. Example: 'search: Find Wikipedia articles by keyword. Returns up to 10 article titles. Use this to explore topics before requesting summaries.'
• Add descriptions to all parameters. Example: 'query (string, 1-200 chars): The search term or article title. Use natural language; e.g., "climate change" or "Albert Einstein". Special characters are converted to spaces.'
• Document output schemas for all tools. Example: 'search returns: {"results": [{"title": str, "snippet": str, "url": str}], "total": int, "limit": int}'. Include field types and what they represent.
• Implement result limits and pagination. Add an optional 'limit' parameter (default 10, max 50) to search(). For page() and summary(), document that full article text is returned and advise agents to trim responses.
• Add error handling and recovery guidance. Catch wikipedia.exceptions.DisambiguationError and return a list of candidates: 'Disambiguation: "apple" could refer to: Apple Inc., apple (fruit), Apple Records. Retry with a more specific query.' Catch PageError and suggest: 'Article not found. Try search() first to find the exact title.'
• Make set_lang idempotent and observable. Return the previous language so agents can restore it if needed: '{"previous_lang": "en", "current_lang": "es"}'.
• Add a 'characters_limit' optional parameter to summary() and page() so agents can request summaries (e.g., first 500 chars) instead of full articles.