If you're a developer looking to deal with MediaWiki's wikitext output, but you would much prefer it to be an HTML DOM, then Parsoid can best help you through its HTTP API that serves HTML (or JSON) responses.

This guide may become out of date; the latest information should be available on the wiki.

/{wiki domain}/v3/page/{html|pagebundle}/{article name}[/{revision}]

GET

If you make a GET request to the API, to a URI that represents a valid wikimedia domain and an article name, you will get back an HTML document with a bunch of extra information used for round-tripping. You can use this to do basic parsing of existing wiki pages.

Responses

Assuming all goes well, you will receive a 200 OK response with the text of the HTML document. But what if things go wrong?

401 Unauthorized

This means that the wiki you were trying to access doesn't allow read access to anonymous users. Parsoid will never work for this wiki, and you should report this as a bug in the server's configuration.

404 Not Found

This means what you would expect - the page was not there on the wiki. It might also indicate that this is a redirect to a different wiki - those do exist - and that you should use the URL in the text of the response instead.

500 Internal Server Error

The least helpful of error codes. We do try to include more information in the body of the response, but it may not always be as helpful as we intend.

/{wiki domain}/v3/transform/{wikitext|html|pagebundle}/to/{wikitext|html|pagebundle}[/{article name}[/{revision}]]

POST

Converts wikitext to html, or vice-versa.

/_version/

GET

Yields a JSON object of the daemon name and version from package.json. If running from a git repository, it would add the sha of the HEAD commit (git rev-parse HEAD). Example:

$ curl http://localhost:8000/_version
{"name":"mediawiki-parsoid","version":"0.0.1","sha":"63a778a1ffc1e9bd0dbb3a7571fe40bfb0a6d699"}