Wikipedia MCP

Give an MCP client search, article-extract, and summary access to English Wikipedia through Wikimedia’s public APIs. Responses are compact JSON serialized inside MCP text content, making the service useful for topic discovery, grounding drafts, entity lookup, and retrieving plain-text encyclopedia context without scraping article HTML.

Use cases

• search English Wikipedia and pass candidate titles to an agent;
• retrieve a plain-text article extract for reading or summarization;
• fetch a concise summary, short description, canonical page URL, and optional thumbnail;
• resolve ambiguous terms before deeper research.

Wikipedia is a community-edited tertiary source. Verify important claims against the references cited by the relevant article and other authoritative sources.

Connect over Streamable HTTP

Use this exact MCP endpoint:

https://reverberant-equality--wikipedia-mcp.apify.actor/mcp

Authenticate every request through Apify:

Authorization: Bearer $APIFY_API_TOKEN
Content-Type: application/json
Accept: application/json, text/event-stream

Configure your client for Streamable HTTP and complete the MCP initialize exchange before tools/list or tools/call.

Tools, arguments, and results

search-wikipedia

• query — required string passed to Wikipedia search.
• maxResults — optional number; default 10. The tool does not impose a local minimum, maximum, or integer constraint, so callers should supply a positive integer accepted by Wikimedia.

Returns a JSON array. Each item contains title, numeric pageId, snippet, and canonical English Wikipedia url. Search snippets can contain Wikimedia’s <span class="searchmatch"> highlighting HTML.

Search first calls the MediaWiki REST search endpoint. If that request fails for any reason, it retries through the MediaWiki Action API.

get-article

• title — required Wikipedia article-title string.

Returns title, plain-text extract, numeric pageId, and article url. It uses the MediaWiki Action API’s prop=extracts response; it does not return page HTML, wikitext, citations as structured objects, revision metadata, or edit history.

get-summary

• title — required Wikipedia article-title string.

Returns title, description, summary extract, thumbnail URL or null, and canonical url. It uses Wikimedia’s page-summary REST endpoint.

Example tool-call payload after initialization:

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get-summary",
    "arguments": {
      "title": "Apollo 11"
    }
  }
}

Illustrative response shape, with the extract shortened:

{
  "title": "Apollo 11",
  "description": "1969 crewed spaceflight that first landed humans on the Moon",
  "extract": "Apollo 11 was a spaceflight...",
  "thumbnail": "https://upload.wikimedia.org/example.jpg",
  "url": "https://en.wikipedia.org/wiki/Apollo_11"
}

Pricing

Each invocation of any tool triggers one tool-call event at $0.005 per call. Charging occurs before the Wikimedia request, so not-found responses, rate limits, and upstream failures may still be charged. Apify platform usage can be billed separately under the caller’s plan.

Upstream sources and error behavior

The service reads only English Wikipedia and uses these Wikimedia interfaces:

• https://en.wikipedia.org/w/rest.php/v1/search/page for primary search;
• https://en.wikipedia.org/w/api.php for fallback search and article extracts;
• https://en.wikipedia.org/api/rest_v1/page/summary/{title} for summaries.

Requests identify the automated client and ask for JSON. Search returns an empty array when Wikimedia reports no matches. get-article and get-summary return JSON such as {"error":"Article \"...\" not found"} with MCP isError: true when a title is missing. Other upstream/network failures return an error text item with isError: true. A primary search failure is not exposed if the Action API fallback succeeds.

Results depend on current Wikimedia content, availability, redirects, disambiguation, and rate limits. Only English Wikipedia is supported. The tools do not choose among ambiguous pages, retrieve arbitrary languages, expose article references as structured citations, or guarantee completeness or factual accuracy. Summary descriptions and thumbnails can be empty or null; search snippets may contain HTML.

Queries and titles are sent to Wikimedia. The implementation does not write responses to an Apify dataset or other application storage, and its normal logs record tool activity without the query text. Do not send confidential terms: upstream and platform infrastructure still process each request.

Local development

Requires Node.js 20+ and the Apify CLI:

npm ci
npm test
npm run build
apify validate-schema
npm run start:dev

The local Streamable HTTP endpoint is http://localhost:3000/mcp.