Article Extractor: News & Blog Text to Markdown
Pricing
from $0.70 / 1,000 articles
Extract news and blog articles from any URL, or find the newest via a site's RSS/Atom feed or sitemap. One clean row each: title, author, published and modified dates, lead image, language, main text as plain text and Markdown, word count, canonical URL. Only-new-articles mode for monitoring.
Versions follow MAJOR.MINOR.PATCH (src/version.py); Apify shows MAJOR.MINOR from .actor/actor.json.
Every run logs its version and records it in the RUN_STATS key-value record.
About future changes in output: main-content detection is heuristic, and sites redesign without notice. The run log and
RUN_STATSname every input and article that failed and why; one broken site never affects the others.
RUN_STATS keeps the reason as before.humble-echidna-<name>, and if that is
refused too it runs without memory for that run. Users' existing memory is untouched: the usual name is tried first.Retry-After answers: when a site or API asks it to wait before retrying, it now waits up
to 90 seconds (was 30), once per site, with at most 3 minutes of such long waits per run; a longer wait, or a
site that asks again after it was waited out, is reported with the reason instead of stalling the run.Chain it after another actor, e.g. RSS & Atom Feed Reader or a news scraper: the full text of every article it found.
datasetId, a dataset picker with read access) and Field with
the article URL (datasetUrlField). With a dataset, each item's link is extracted like an article URL, and
sites and articleUrls are ignored, defaults included: a run started from an integration (which gets the input
schema's defaults filled in) never fetches or charges for the Rust blog. The field is found automatically when left
empty: the first of url (RSS & Atom Feed Reader's and GDELT News Search's link), link, articleUrl with a web
address in the first 100 items, then the same names one level down (article.link); a dotted path names a nested
one.utm_*, gclid, fbclid, ...) are dropped. All
counted in the new RUN_STATS.dataset record (itemsRead, websites (the article URLs kept), withoutWebsite,
notAWebsite, duplicates, the field used and whether it was found automatically). A dataset with no article URL
at all, or one that can't be read, fails the run with the reason.sourceTitle (the item's title, else name) and sourceIndex (its
position in the dataset, 0 = the first), to join articles back to the items; null for articles not read from a
dataset. Such articles have discoveredVia: "dataset" and the dataset id in discoveredFrom.sites and articleUrls say that a left-out sites takes its default (the Rust blog), so an agent extracting only
its own article URLs passes "sites": []; the README's MCP answer says the same.First release.
Sitemap: lines, then the usual locations;
indexes and news sitemaps included). Newest first, up to "Max articles per site"; URL patterns keep sitemaps to
articles; undated sitemap entries are ordered by a date in their URL (/2026/09/22/), never output as one;
discovered pages under 50 words (tag pages, listings) are skipped and not charged.@graph included), Open Graph / article:*, common publisher meta tags,
microdata and <time>; the feed or news sitemap fills in title, author and date when the page doesn't say.mms_extract, mms_feeds, mms_sitemap), not copies.apify-default-dataset-item event. "Max articles per run" and the
maximum cost per run are honoured before each article is fetched; a failed article gives its place back.