Wikipedia MCP Server

Free knowledge access for AI agents. Search, read, and explore 6.8M+ Wikipedia articles with zero API keys. Built for agents that need facts, summaries, and discovery.

Agent: "What's the capital of Bhutan?"
  → wiki_get_summary("Thimphu")
  → "Thimphu is the capital and largest city of Bhutan..."

Agent: "What are the categories of 'Quantum computing'?"
  → wiki_get_categories("Quantum computing")
  → ["Quantum information science", "Computational complexity theory", ...]

Architecture

┌──────────────┐     stdio/JSON-RPC      ┌──────────────────┐
│  AI Agent     │ ◄──────────────────────► │  Wikipedia MCP   │
│  (Claude, etc)│                         │  Server (Python)  │
└──────────────┘                         └───────┬────────────┘
                                                  │  HTTPS
                                                  ▼
                                         ┌──────────────────┐
                                         │  Wikipedia REST  │
                                         │  API (Free, No   │
                                         │  Auth Required)   │
                                         └──────────────────┘

Why Wikipedia? 6.8M+ English articles, CC BY-SA 4.0 licensed, comprehensive, constantly updated. No API key, generous rate limits (~200 req/s), and 300+ language editions.

Tools

| # | Tool | Description | Parameters | |---|------|-------------|-----------| | 1 | wiki_search | Search Wikipedia by query | query (required), limit, language, response_format | | 2 | wiki_get_article | Get full article content | title (required), language, max_sections, response_format | | 3 | wiki_get_summary | Get short summary/extract | title (required), sentences, language, response_format | | 4 | wiki_get_categories | Get article categories | title (required), limit, language, response_format | | 5 | wiki_get_languages | Get available translations | title (required), limit, response_format | | 6 | wiki_random | Get a random article | language, response_format | | 7 | wiki_page_info | Get page metadata | title (required), language, response_format |

Tool Details

wiki_search — Find articles

Search by natural language query. Returns titles, snippets (with bold highlights), page IDs, word counts, and direct URLs. Supports 300+ language editions.

// Example: wiki_search("quantum entanglement", limit=5)
{
  "query": "quantum entanglement",
  "total_hits": 1234,
  "results": [
    {
      "title": "Quantum entanglement",
      "pageid": 24934,
      "snippet": "**Quantum entanglement** is a physical phenomenon...",
      "word_count": 15623,
      "url": "https://en.wikipedia.org/wiki/Quantum_entanglement"
    }
  ]
}

wiki_get_article — Read full articles

Returns the complete article text as markdown with section headings. Use max_sections to limit for very large articles (>50KB). All content is CC BY-SA 4.0 licensed.

// Example: wiki_get_article("Python (programming language)", max_sections=3)
{
  "title": "Python (programming language)",
  "length_chars": 42310,
  "extract": "# Python (programming language)\n\n**Python** is a high-level...",
  "categories": ["Programming languages", "Python (programming language)"],
  "url": "https://en.wikipedia.org/wiki/Python_(programming_language)"
}

wiki_get_summary — Quick facts

Returns the introductory extract (2-5 sentences by default) plus key metadata. Perfect for quick fact-checking, trivia, and agent decision-making.

wiki_get_categories — Navigate the knowledge graph

Returns all categories an article belongs to. Filters out Wikipedia infrastructure categories. Useful for topic exploration and building knowledge taxonomies.

wiki_get_languages — Cross-language access

Shows all available language versions of an article with language codes, native names, and full URLs. Supports 300+ languages from Afrikaans to Zulu.

wiki_random — Serendipitous discovery

Returns a random article with summary. Great for icebreakers, exploration, and "did you know?" moments.

wiki_page_info — Metadata at a glance

Returns page size, last modified date, content model, redirect status, and preview text.

Quality Standards

All tools implement Anthropic's MCP quality standards:

| Standard | Implementation | |----------|---------------| | Tool Annotations | All 4 booleans: readOnlyHint, destructiveHint, idempotentHint, openWorldHint | | Service Prefix | wiki_ prefix on all tools to avoid namespace collisions | | Dual Response Format | markdown (human-readable) or json (programmatic) | | Pagination | limit parameter on all list tools | | CHARACTER_LIMIT | 25,000 char limit with automatic truncation + guidance | | Error as Result | Errors returned as JSON with isError: true and next_steps |

Installation

Prerequisites

• Python 3.10+
• pip

Setup

git clone https://github.com/Rumblingb/wikipedia-mcp.git
cd wikipedia-mcp
pip install -r requirements.txt

Configure in Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "wikipedia": {
      "command": "python3",
      "args": ["server.py"],
      "cwd": "/path/to/wikipedia-mcp"
    }
  }
}

Configure in VS Code / Cursor

{
  "mcpServers": {
    "wikipedia": {
      "command": "python3",
      "args": ["server.py"],
      "cwd": "/path/to/wikipedia-mcp"
    }
  }
}

Deploy to Smithery

Visit smithery.ai → Import from GitHub → Select wikipedia-mcp.

Pricing

| Tier | Price | Queries/Month | Support | |------|-------|---------------|---------| | Free | $0 | 50 | Community | | Pro | $19/mo | Unlimited | Priority Email | | Enterprise | $99/mo | Unlimited + SLA | Dedicated |

👉 Subscribe to Pro →

FAQ

Q: Is this really free? No API key? Yes. Wikipedia's API is public and requires no authentication. The free tier gives you 50 queries/month through this server.

Q: What about content licensing? All Wikipedia content is licensed under CC BY-SA 4.0. Attribution is included in article responses.

Q: Which languages are supported? 300+ language editions. Use the language parameter on any tool (e.g., language="es" for Spanish).

Q: How is this different from just hitting the Wikipedia API directly? This server adds: agent-friendly response formats (markdown + JSON), automatic truncation, category filtering, error recovery with actionable next steps, and MCP protocol compliance for direct agent integration.

Q: Can I contribute? Yes! PRs welcome. See CONTRIBUTING.md.

Developer Notes

• Rate Limiting: Wikipedia allows ~200 requests/second. Be respectful — add delays between requests if making many calls.
• Attribution: Always attribute Wikipedia content per CC BY-SA 4.0.
• Caching: Consider caching results for repeated queries — Wikipedia content rarely changes.

Built by AgentPay Labs — Governed payment middleware for AI agents.