MediaWiki Search API
Wikimedia FoundationThe search surface of the MediaWiki Action API, served at the `api.php` endpoint of every MediaWiki site, including Wikipedia and Wikinews, which answer it for anonymous callers. Its `list=search` module returns ranked page titles with highlighted snippets and page metadata, and can run as a generator that passes its results to the TextExtracts module, which returns plain-text extracts of the pages found. On Wikimedia's wikis it is backed by the CirrusSearch extension over an OpenSearch index.
Scored as the composable search surface of MediaWiki's Action API, not as Wikipedia: `list=search` and `generator=search` with the TextExtracts module, which has shipped bundled with MediaWiki since 1.34, and CirrusSearch, the backend Wikimedia deploys. Both GitHub repositories are mirrors of Wikimedia's own Gerrit. TextExtracts' documentation asks Wikimedia's own production features to use the Page Content Service for summaries instead. Wikimedia sets its own API rate limits, and MediaWiki's guidance asks callers to send requests in series rather than in parallel.
Openness
5 high confidence- license
- GPL-2.0(OSI
- source
- public(MediaWiki core and the CirrusSearch and TextExtracts extensions, over an OpenSearch index)
- self-host
- yes(any MediaWiki install can enable CirrusSearch against its own OpenSearch cluster)
- core features withheld
- no — the Wikimedia Foundation runs it as a free public API and describes the engine as open source
MediaWiki and the CirrusSearch extension behind its search are both licensed GPL-2.0-or-later, and the extension's README walks through running the same search against your own OpenSearch cluster. The Wikimedia Foundation's search team describes what it runs as an open-source search engine, and nothing in the API is sold, so there is no paid tier for anything to be held back for.
- https://raw.githubusercontent.com/wikimedia/mediawiki-extensions-CirrusSearch/HEAD/COPYING recorded 2026-10-13
The COPYING body: "GNU GENERAL PUBLIC LICENSE Version 2, June 1991", full text.
- https://raw.githubusercontent.com/wikimedia/mediawiki-extensions-CirrusSearch/HEAD/extension.json recorded 2026-10-13
The extension manifest declares "license-name": "GPL-2.0-or-later".
- https://raw.githubusercontent.com/wikimedia/mediawiki-extensions-CirrusSearch/HEAD/README recorded 2026-10-13
Installation: "Get OpenSearch v2.x or v3.x up and running somewhere", load the Elastica and CirrusSearch extensions, build the index with the maintenance scripts, then set `$wgSearchType = 'CirrusSearch'`.
- https://raw.githubusercontent.com/wikimedia/mediawiki-extensions-TextExtracts/master/COPYING recorded 2026-10-13
The COPYING file is the "GNU GENERAL PUBLIC LICENSE Version 2, June 1991" text.
- https://raw.githubusercontent.com/wikimedia/mediawiki-extensions-TextExtracts/master/extension.json recorded 2026-10-13
TextExtracts' extension.json gives "name": "TextExtracts" and "license-name": "GPL-2.0-or-later", the same license as MediaWiki core and CirrusSearch.
- https://raw.githubusercontent.com/wikimedia/mediawiki/HEAD/COPYING recorded 2026-10-13
"MediaWiki is licensed under the terms of the GNU General Public License, version 2 or later."
- https://www.mediawiki.org/w/index.php?title=Wikimedia_Search_Platform&action=raw recorded 2026-10-13
The page's wikitext: the Search Platform team maintains "the CirrusSearch extension which relies on OpenSearch (formerly Elasticsearch), the search backend used at the Wikimedia Foundation to support Wikimedia projects", and: "We provide an open-source search engine, backed by an inverted index for non-structured on-wiki data."
Adoption
not assessedWikimedia publishes no request count for the search module, or for the Action API as a whole, on the pages read. The API needs no key, account or client package, so there is no sign-up or download figure to count instead either. No level is recorded.
- https://www.mediawiki.org/w/index.php?title=API:Search&action=raw recorded 2026-10-13
The wikitext of the module's reference page: its help template, an example GET request and response, sample code and parameter history, with no request or user count, which is why no level is recorded.
Capability
4 medium confidenceA search returns pages ranked by relevance with highlighted snippets, and the same query can run as a generator so that one call also returns a plain-text extract of each page found. Finding pages and returning their cleaned text through one interface puts it level with DDGS. It searches the wiki it is called on rather than the open web, which limits what it covers rather than what it does, and it composes no answer.
- https://en.wikipedia.org/w/api.php?action=paraminfo&modules=query%2Bextracts&format=json recorded 2026-10-13
Wikipedia's paraminfo for `query+extracts` names its source as "TextExtracts", licensed "GPL-2.0-or-later", with parameters including `chars`, `sentences`, `intro` and `plaintext`, so the module is live on the wiki Alpha searches.
- https://en.wikipedia.org/w/api.php?action=paraminfo&modules=query%2Bextracts&helpformat=wikitext&format=json recorded 2026-10-13
English Wikipedia's description of `prop=extracts`: source TextExtracts, license GPL-2.0-or-later, "Returns plain-text or limited HTML extracts of the given pages", with `explaintext`: "Return extracts as plain text instead of limited HTML."
- https://en.wikipedia.org/w/api.php?action=paraminfo&modules=query%2Bsearch&helpformat=wikitext&format=json recorded 2026-10-13
English Wikipedia's own description of `list=search`: module "search", source MediaWiki, license GPL-2.0-or-later, "Perform a full text search." `srsearch` "Search for page titles or content matching this value"; `srsort` default "relevance"; `srqiprofile` ranking profiles including classic, popular_inclinks_pv and mlr-1024rs; `srprop` default "size|wordcount|timestamp|snippet". Examples include `action=query&generator=search&gsrsearch=meaning&prop=info`.
- https://www.mediawiki.org/w/index.php?title=Extension:TextExtracts&action=raw recorded 2026-10-13
The TextExtracts extension page: it "provides an API which allows retrieval of plain-text or limited HTML" extracts of page content, the `prop=extracts` module "Returns plain-text or limited HTML extracts of the given pages" with an `explaintext` parameter, and the page carries "{{Bundled|1.34}}". It also asks that new features for code running in Wikimedia production use the Page Content Service for summaries instead.
Verified 2026-10-13