Wikipedia API

The Wikipedia API returns article summaries as structured rows — title, page id, the short Wikidata description, the lead-section extract, canonical URL, thumbnail, last-modified time and the Wikidata id. Search with a query, or pass an exact title list or /wiki/ URLs and read those articles directly.

It delivers the lead — the summary you would read at the top of an article — not the full body text, which is the right granularity for enrichment, entity cards and knowledge panels. Any language edition works through the lang input, so the same title can be pulled in English, Italian or German. Titles that do not exist come back under failed and are never billed. It reads Wikipedia's own summary endpoint over the TLS tier — no browser.

$0.0005 per delivered article, up to 50 articles per run. Nothing delivered means nothing charged.

How the Wikipedia API works

You either send a query to search and read the top results, or a titles array of exact article titles or /wiki/ URLs to read directly — one or the other. Set lang for the edition and max_results up to 50. Each article becomes one flat row.

Two things worth knowing:

Inputs

Use query to search Wikipedia and read the top hits, or titles to read exact articles or /wiki/ URLs — one or the other, not both. lang picks the language edition (default en). max_results caps delivery at 50 articles.

What one article looks like

Each article is one flat row: title, page_id, the short description, the lead-section extract, canonical url, thumbnail, the lang read, an ISO last_modified and the wikidata_id (Q-number). That Wikidata id is the join key to structured facts elsewhere, which is why it is a first-class field rather than something you resolve separately.

What the Wikipedia API costs

$0.0005 per delivered article ($0.50 per 1,000). Nothing delivered means nothing charged, and the $2 monthly free credit covers roughly 4,000 articles here. Volume tiers take up to 30% off.

$0.0005 per delivered article, $0.50 per 1,000. Enriching a 10,000-entity list with a summary and Wikidata id apiece is $5. Titles that do not resolve are returned under failed and cost nothing, so a messy input list degrades coverage rather than inflating the bill.

The $2 monthly allowance (no card) covers about 4,000 delivered articles, and volume tiers take up to 30% off. GET /v1/scraper/collectors returns the exact price your key pays.

Wikipedia API vs the MediaWiki API

Being straight about it: Wikipedia's own MediaWiki and REST APIs are free, official and well-documented, and this collector reads that same summary endpoint. If Wikipedia is your only source and your volume is modest, calling it directly is a sound choice — there is no access barrier to get around.

What this adds is shape and scale. The raw MediaWiki API is powerful but sprawling — several modules, continuation tokens, and responses you assemble yourself. This returns one flat, typed row per article with the extract, description, thumbnail and wikidata_id already lined up, a failed bucket for bad titles, and delivery under the same QuantumProxies key and per-row billing as your other collectors — with rotating exits so a big enrichment batch is not throttled by one IP.

Versus calling MediaWiki yourself

MediaWiki's API is open, so the DIY cost is assembly and operation rather than access:

If low-volume Wikipedia lookups are the whole job, the official API may be enough; at enrichment scale or alongside other sources, this and our residential proxies are the operational win.

What people build with it

Entity enrichment

Feed a list of people, companies or places as titles and get a summary, thumbnail and wikidata_id for each — enough to build a knowledge card or join to structured data downstream.

Knowledge panels and tooltips

Use the lead extract and description to power hover cards or panels in your own product, refreshed as articles change via last_modified.

Multilingual reference

Pull the same title across lang editions to compare how a topic is summarised by market, or to localise an entity card without a separate translation step.

Limits, scope and the legal bit

One run returns up to 50 articles. The extract is the lead-section summary, not the full article body — this collector is built for summaries and enrichment, not for mirroring page content. Non-existent titles are returned under failed, unbilled, never as empty rows. Any language edition is available through lang. Runs parallelise up to your plan's rate limit — 60 a minute on pay-as-you-go, up to 1,200 on the top tier.

Legally: Wikipedia content is public and licensed (generally CC BY-SA), so reuse carries attribution and share-alike obligations you should honour; the MediaWiki endpoint also has its own terms. Background, not legal advice.

FAQ

Is the Wikipedia API free?

$2 of usage a month with no card, which at $0.0005 per article is about 4,000 delivered articles. Wikipedia's own MediaWiki API is free too; this collector is worth it mainly when you want flat typed rows, a failed bucket for bad titles, and delivery alongside your other collectors.

Do I need an API key for Wikipedia?

No. Your QuantumProxies API key is the only credential — the MediaWiki API needs none either, and this collector reads it for you, so there is nothing extra to register or rotate.

Does it return the full article text?

No — it returns the lead-section extract, the opening summary, plus the short description, thumbnail and Wikidata id. That is the enrichment-sized slice; mirroring full article bodies is a different job with different licensing weight.

What happens to titles that do not exist?

They are returned under failed and are not billed, rather than coming back as a row with an empty extract. That keeps a messy input list from silently corrupting your dataset or padding your invoice.

Related scrapers