Wikipedia Scraper

Search Wikipedia by keyword, or fetch clean, structured page data for exact titles — straight from the official MediaWiki Action API. No API key, no login, no anti-bot.

Two modes

1. Search — set searchQuery. Returns matching articles with title, pageid, url, a plain-text snippet (the API's HTML is stripped for you), wordcount, size, and timestamp. The actor paginates automatically (50 per request) up to maxItems.

2. Page data — set pageTitles (a list of exact article titles). Returns title, pageid, url, the plain-text extract, a thumbnail image URL, and categories. Titles are batched 50 at a time. Turn on fullText to get the whole article instead of just the intro.

(If both are provided, search mode wins. Provide one or the other.)

What you get per row

Field	Mode	Notes
title	both	Article title.
pageid	both	Stable Wikipedia page id (used to dedupe).
url	both	Canonical article URL.
snippet	search	Plain-text match snippet (HTML stripped).
wordcount, size, timestamp	search	Article word count, byte size, last-edit time.
extract	page	Plain-text article text (intro, or full body with fullText).
thumbnail	page	Lead image URL (up to 400px), if the page has one.
categories	page	Visible category names (hidden categories excluded).

Input

Field	Notes
searchQuery	Keywords, e.g. machine learning. Leave empty if using titles.
pageTitles	List of exact titles, e.g. ["Apify", "Web scraping"].
fullText	Page mode only. Full article text vs. just the intro. Default off.
language	Wikipedia edition: en, fr, de, es, ja, … Default en.
maxItems	Cap on returned pages. Default 50.

Output

One dataset row per page (ok: true), deduplicated by pageid. Empty searches or unknown titles return a non-charged diagnostic row with an errorCode and a human-readable reason instead of silently returning nothing.

Pricing

$1.00 per 1,000 pages ($0.001 each), with no run-start fee. Flat rate — no volume tiers, no plan gates — and you are charged only for pages actually returned.

A row is one article, in either mode, whether you asked for the intro or the full body with fullText. Long articles cost exactly the same as short ones. A search that matches nothing, or a title Wikipedia does not have, produces an uncharged diagnostic row — so a run that finds nothing costs nothing.

What people use it for

• RAG and LLM corpora — pageTitles with fullText: true gives clean plain-text article bodies with the wiki markup, references and infobox clutter already stripped, 50 titles per batch.
• Entity enrichment — resolve a list of company, person or place names to canonical titles, pageids and lead images to attach to your own records.
• Topic mapping — categories on page rows lets you build a subject graph without parsing category pages yourself.
• Cross-language lookups — the same title list against language: "fr", "de", "ja" and so on, since every Wikipedia edition speaks the same API.
• Change tracking — search mode returns timestamp (last edit) and size, so you can re-run and diff which articles moved.

Example

{ "searchQuery": "machine learning", "language": "en", "maxItems": 30 }

{ "pageTitles": ["Apify", "Web scraping"], "fullText": false, "language": "en" }

Notes

Uses https://{language}.wikipedia.org/w/api.php. Per Wikimedia's policy the actor always sends a descriptive User-Agent with a contact. Results are deduped by pageid.