Search Wikipedia articles, read summaries and full text, target sections, find nearby pages, and list language editions via @cyanheads/wikipedia-mcp-server (npx) — 6 tools

@cyanheads/wikipedia-mcp-server v0.1.10 — 6 tools for Wikipedia

Install & run: npm install @cyanheads/wikipedia-mcp-server → entry dist/index.js (ESM, type: module)

Tools

Key gotchas

1. section_index starts at 1 — passing 0 returns a clear error with recovery hint; omit section_index entirely to get the lead section + full article
2. language defaults to `en` — pass any Wikipedia language code (e.g. tr, fr, de, ja) for non-English editions; search, summaries, and articles all support it
3. Nonexistent titles return structured errors (not exceptions) with recovery hints pointing to wikipedia_search
4. Summaries include Wikidata QID (wikibase_item field, e.g. Q406 for Istanbul) — useful for cross-referencing with Wikidata/structured knowledge graphs
5. Nearby search returns 0 results gracefully for remote locations (tested Sahara desert at 10km radius — empty array, no error)
6. Language editions can be extensive — Istanbul has 230 language editions; Python (programming language) has 118
7. Search returns word counts per article — useful for estimating article depth before fetching full text
8. First call ~1000ms (geocoding + API), subsequent calls 70-500ms depending on response size
9. No authentication needed — uses Wikipedia's public REST API
10. Disambiguation handled by Wikipedia — searching "Model Context Protocol AI" correctly finds the MCP Wikipedia article

Verified trace (14 calls, 100% success)

✓ search("Model Context Protocol AI", limit=3) → 3 results, top: "Model Context Protocol" (1230 words) — 1029ms
✓ search("Istanbul Turkey", limit=3) → 3 results, top: "Istanbul" (23796 words) — 477ms
✓ get_summary("Istanbul") → "Istanbul is the largest city in Turkey..." + QID Q406 + thumbnail — 106ms
✓ get_summary("Rust (programming language)") → "Rust is a general-purpose programming language..." + QID Q575650 — 133ms
✓ get_sections("Istanbul") → 44 sections with indices/levels/numbering — 283ms
✓ get_article("Istanbul", section_index=1) → Names section, 2040 chars, Byzantium history — 1072ms
✓ search_nearby(41.0082, 28.9784, 500m, limit=5) → 5 articles: "740 Constantinople earthquake" (0m), "Milion" (37m), "Basilica Cistern" (54m) — 396ms
✓ get_languages("Istanbul") → 230 language editions — 373ms
✓ search("yapay zeka", language="tr", limit=3) → 3 Turkish results: "Yapay zekâ" (4897 words) — 525ms
✓ get_summary("İstanbul", language="tr") → Turkish summary + QID Q406 — 227ms
✓ get_article("Rust...", section_index=0) → Error: "indices start at 1" + recovery hint — 14ms (correct rejection)
✓ get_summary("Xyzzy_nonexistent_article_12345") → Error: "No article found" + recovery hint — 450ms (correct rejection)
✓ get_languages("Python (programming language)") → 118 language editions — 356ms
✓ search_nearby(23.42, 25.66, 10000m) → 0 articles (Sahara desert) — 438ms (correct empty result)

Latencies: min=14ms, p50=396ms, max=1072ms. First geocoded search ~1s, steady-state 70-500ms.