search_wikipedia and wikipedia_search_wikipedia are the same tool under two names, and that doubling runs through the whole list; it costs context on every turn without adding anything a caller can do. What the distinct set offers is genuinely worth having: summarize_article_for_query cuts around your question instead of returning the lead paragraph, and summarize_article_section, get_sections and extract_key_facts let a caller take a page apart rather than swallow it whole. test_wikipedia_connectivity is the first thing to reach for when results look wrong, since it separates a network failure from an empty answer and reports response time with it.
A Wikipedia reader built to return fragments rather than whole articles: query-scoped and section-scoped summaries, fact extraction, related topics, section and link listings, coordinates, and an API connectivity check.
- search_wikipedia returns articles matching a query, cut to limit.
- get_article returns an article's full content and get_summary returns its summary alone.
- summarize_article_for_query returns a snippet from around your query inside the article, sized by max_length, so a long page can answer a narrow question.
- summarize_article_section summarises the section named in section_title instead of the whole page.
- extract_key_facts returns a list of facts from an article, as many as count asks for, optionally narrowed by topic_within_article.
- get_related_topics derives neighbouring topics from an article's links and categories, bounded by limit.
- get_sections and get_links return an article's section list and its outgoing links, and get_coordinates returns its coordinate information.
- test_wikipedia_connectivity reports the base API URL, the language, site information and response time in milliseconds, or a failed status with the error attached.
No credentials are configured; calls go to the Wikipedia API for whichever language the server is set to, which test_wikipedia_connectivity reports back along with the base API URL.
One command — uvx wikipedia-mcp