Yes. The server is open source under the MIT license and free to run, and Wikipedia's API is free to query. No key is required for basic use; an optional access token only raises rate limits.

Does it support remote/OAuth?

By default it runs locally over stdio (via uvx), and there is no hosted OAuth endpoint. It can optionally be served over HTTP/SSE transports if you want to host it yourself, but authentication is not required to read Wikipedia.

Can it use non-English Wikipedia?

Yes. Pass the --language flag (for example ja or zh-hans) to target a specific Wikipedia language edition, with an optional --country flag, so the same server can ground answers in the language you need.

Wikipedia MCP server

CommunityRavi Kumar E257Config last verified Jun 1, 2026

A maintained MCP server that grounds an agent in Wikipedia: search articles, pull full text or summaries, sections, links, and key facts.

This Wikipedia MCP server is an actively maintained, community-built server that lets an AI agent retrieve information from Wikipedia to ground its answers in a citable source. Rather than relying on memorized (and potentially stale) facts, the agent can search Wikipedia for matching articles, fetch the full text of an article or a concise summary, list an article's sections or outbound links, get an article's geographic coordinates, and pull related topics. It also offers higher-level helpers: a query-tailored summary, a section-specific summary, and a key-facts extractor, so the agent can get exactly the slice of an article it needs instead of dumping an entire page into context. Every tool is also exposed with a wikipedia_ prefix to avoid name collisions with other servers.

The server is published as the wikipedia-mcp Python package and runs locally over stdio, launched with uvx (uvx wikipedia-mcp) or installed via pip/pipx. It supports any Wikipedia language edition through a --language flag (and a --country option), optional response caching with --enable-cache, and an optional WIKIPEDIA_ACCESS_TOKEN to lift rate limits. It can also run over HTTP/SSE transports if you prefer to host it. No key is required for basic use, making it a low-friction factual-grounding tool for research, fact-checking, and RAG-style workflows.

Quick install

Copy-paste configs are provided for all 8 supported clients. Pick your client below.

Add to ~/.claude.json

~/.claude.json

json

{
  "mcpServers": {
    "wikipedia": {
      "command": "uvx",
      "args": [
        "wikipedia-mcp"
      ]
    }
  }
}

Or via CLI

bash

claude mcp add wikipedia -- uvx wikipedia-mcp

Step-by-step guides

Available tools

Tool	Description
search_wikipedia	Searches Wikipedia for articles matching a query.
get_article	Gets the full content of a Wikipedia article.
get_summary	Gets a concise summary of a Wikipedia article.
get_sections	Gets the sections of a Wikipedia article.
get_links	Gets the links contained within a Wikipedia article.
get_coordinates	Gets the geographic coordinates of a Wikipedia article.
get_related_topics	Gets topics related to a Wikipedia article.
summarize_article_for_query	Gets a summary of an article tailored to a specific query.
summarize_article_section	Gets a summary of a specific section of an article.
extract_key_facts	Extracts key facts from a Wikipedia article.
test_wikipedia_connectivity	Tests connectivity to the Wikipedia API.

Required configuration

WIKIPEDIA_ACCESS_TOKENOptional
Wikimedia personal access token to lift API rate limits. Optional.

What you can do with it

Ground answers in a citable source

The agent searches Wikipedia, pulls a summary or the full article, and uses extract_key_facts to anchor its response in verifiable facts rather than memorized (and possibly stale) knowledge.

Targeted research without flooding context

Use summarize_article_for_query and summarize_article_section to fetch just the relevant slice of a long article, keeping the agent's context focused while still citing the source.

FAQ

Is it free?: Yes. The server is open source under the MIT license and free to run, and Wikipedia's API is free to query. No key is required for basic use; an optional access token only raises rate limits.
Does it support remote/OAuth?: By default it runs locally over stdio (via uvx), and there is no hosted OAuth endpoint. It can optionally be served over HTTP/SSE transports if you want to host it yourself, but authentication is not required to read Wikipedia.
Can it use non-English Wikipedia?: Yes. Pass the --language flag (for example ja or zh-hans) to target a specific Wikipedia language edition, with an optional --country flag, so the same server can ground answers in the language you need.

← Browse all search-and-data servers