Wikipedia MCP

- 🔍Advanced Search— find articles with relevance ranking
- 📄Full Article Access— plain text extracts or complete wikitext
- 📝Smart Summaries— concise summaries tailored to a specific query
- 🗂️Section Analysis— extract and summarize individual sections
- 🔗Link Discovery— internal links and related articles
- 🌍Multi-language— 50+ language codes, 140+ country codes, resolved automatically
- 🔀Parallel Search—multi_search_wikipediaruns multiple queries simultaneously
- 🎯Fact Extraction— structured key-fact lists from any article
- 📍Coordinates— lat/lon for geographic articles
- ⚡Streamable HTTP— full MCP transport support (POST + GET SSE + DELETE)
- 🔒Security hardened— input sanitization, tiered rate limiting, noServerheader
- 🚀Vercel ready— stateless serverless deployment, zero cold-start friction

git clone https://github.com/YOUR_USERNAME/wikipedia-mcp-vercel.git cd wikipedia-mcp-vercel vercel

Or use theDeploy with Vercelbutton above.

The server implementsMCP Streamable HTTP transport(2025-03-26).
Both the POST channel (client → server) and the GET SSE channel (server → client) are supported.

MCP endpoint:https://your-app.vercel.app/mcp

{ "mcpServers": { "wikipedia": { "type": "http", "url": "https://your-app.vercel.app/mcp" } } }

claude mcp add wikipedia --transport http https://your-app.vercel.app/mcp

Configure the client with transport typestreamable-httpand the endpoint URL above.
The server usesstateless mode(noMcp-Session-Idheader), which is correct and expected for serverless deployments.

Copy.env.example→.envfor local development.

# Health check curl https://your-app.vercel.app/health # Test MCP initialize curl -X POST https://your-app.vercel.app/mcp \ -H "Content-Type: application/json" \ -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-03-26","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}'

All 13 tools accept optionallanguage(e.g."ja") andcountry(e.g."JP") parameters. When set, the query is sent as-is to that language's Wikipedia endpoint (e.g.ja.wikipedia.org). Wikipedia's own search can often match a short English title phrase to the correct local article, but there is no translation layer in this server.

Three prompt templates are registered and available viaprompts/list/prompts/get:

Standard HTTP endpoints for non-MCP usage:

GET /health GET /search/:query?limit=10 GET /article/:title GET /summary/:title GET /sections/:title GET /links/:title GET /coordinates/:title GET /related/:title?limit=10 GET /summary/:title/query/:query/length/:maxLength GET /summary/:title/section/:section/length/:maxLength GET /facts/:title?topic=...&count=5 GET /test-connectivity GET /supported-countries POST /tools/:toolName

src/ ├── server.ts # Express app wiring ├── routes/ │ ├── mcp.ts # Streamable HTTP transport (POST + GET SSE + DELETE) │ └── rest.ts # REST endpoints ├── mcp-server.ts # MCPServer helper — tool dispatch, language routing ├── handlers.ts # Tool handler functions ├── toolRegistrations.ts # SDK tool + prompt registration ├── toolDefinitions.ts # JSON Schema definitions (for REST /mcp info) ├── prompts.ts # Prompt template logic ├── wikipedia-client.ts # Wikipedia Action API client ├── sanitize.ts # Input sanitization ├── middleware.ts # Rate limiting, Server header removal ├── utils.ts # Shared helpers └── types.ts # TypeScript interfaces

The server usesstateless Streamable HTTP— the correct mode for serverless platforms:

- POST /mcp— receives JSON-RPC messages, returns JSON or SSE stream
- GET /mcp— opens an SSE channel for server-initiated messages (clients reconnect on timeout)
- DELETE /mcp— session termination (acknowledged; no server-side sessions to clean up)

A freshMcpServer+StreamableHTTPServerTransportinstance is created for each request. All Wikipedia logic is handled by the shared, statelessMCPServerhelper.

npm install npm run dev # tsx watch mode on port 8000 npm run build # tsc → dist/ npm start # node dist/server.js vercel dev # local Vercel simulation

For higher Wikipedia API rate limits, create a bot password:
- Log in to your Wikipedia account
- Go toSpecial:BotPasswords
- Create a new bot password
- SetWIKIPEDIA_BOT_USERNAME=YourAccount@BotNameandWIKIPEDIA_BOT_PASSWORD=...

SerpApi MCP Server for Google and other search engine results