Game wikis, fan wikis and internal MediaWiki installs all speak the same REST API, and pointing this at one turns a niche knowledge base into something a model can search and quote. The repository documents how to check whether a site qualifies: open its base URL followed by /rest.php/v1/page in a browser and see whether the output looks right, adding /w to the base if not. Read the maintainer's notice at the top of the repository first — they consider the project superseded.
A Python server over the MediaWiki REST API that searches a wiki and returns the full text of a page by title.
- Search across a wiki, with a query and a result limit that defaults to 5
- The full content of a page fetched by its exact title
- Any MediaWiki-powered site rather than Wikipedia alone — fandom.com and wiki.gg wikis work once you pass their endpoint as --base-url
- A choice of transport: stdio by default, or streamable HTTP and SSE with --http or --sse and a --port that defaults to 8000
- Proxy support through HTTP_PROXY, for networks where the wiki is not directly reachable
Nothing to authorize — no account, no key. Run with uvx as mediawiki-mcp-server. Note before you install: the repository's own header says it is outdated and points at ProfessionalWiki/MediaWiki-MCP-Server instead.
One command — uvx mediawiki-mcp-server --base-url https://example.com/
