RSS & Atom Keyword Filter

Get matching entries from up to ten public feed URLs in one dataset. Use include and exclude phrases to narrow a release-note, industry-news or supplier-blog watchlist before sending results to your own workflow. No linked articles are fetched and no AI summary is generated.

What it does

• Reads direct RSS or Atom feed URLs; it does not discover feeds from homepages.
• Matches case-insensitive substrings in each entry's title and feed summary. Choose any or all include phrases. An exclude phrase overrides a match. Empty include phrases return all entries that pass the other filters.
• Optionally keeps only entries within maxAgeHours of run start. This uses the feed's published date, falling back to updated date. Undated and future-dated entries are skipped when this filter is enabled; no date accuracy is promised.
• Deduplicates by article URL, or feed-scoped ID when no usable URL exists, within the run. It returns the first matching occurrence across feeds.
• Outputs title, article URL, reported timestamp, matching phrases, stable key and at most 280 characters of the feed summary. SUMMARY records failures, scanned/filtered counts and whether an output or budget limit stopped the run.

Example input

{
  "feedUrls": ["https://blog.cloudflare.com/rss/"],
  "keywords": ["api", "deprecat"],
  "excludeKeywords": [],
  "matchMode": "any",
  "maxAgeHours": 168,
  "maxItems": 50
}

This is an input example, not a promise that the current feed contains a match. An empty dataset may mean no entries survived the filters, or that a budget limit prevented output. Check SUMMARY for feed errors and stopReason; undated exclusions are counted separately.

Pricing

USD0.001 per stored matching item (USD1 per 1,000), with platform usage included and no start fee. Fifty delivered items would cost USD0.05. Filtered entries, duplicates, empty feeds and feed errors do not produce a paid item event.

Limits and use

The tool reads only the entries currently supplied by each feed, at most the first 500 per feed. It fetches at most 2 MB per feed, follows at most five redirects and allows only public HTTP(S) addresses on standard ports. Compressed responses, malformed feeds, login-only feeds and unsupported feed formats are reported as errors. A failing feed does not discard results from other feeds; an all-feed failure marks the run failed with no item charges.

The global output cap is 200 items; default 50. Fetches are sequential, with a 150-second deadline for starting fetches and item writes; an in-progress operation and the final summary may finish later. Network restrictions or publisher rate limits can prevent access. There is no proxy, CAPTCHA solving, paywall bypass or browser.

This Actor does not send alerts or create schedules. You can connect its dataset to an existing integration or set your own Apify schedule. There is no saved cross-run history: overlapping time windows may return and charge the same item again. Use itemKey in your destination to deduplicate across runs. Changes in URLs or source IDs can change that key. Do not use it for exhaustive monitoring or assume that missing entries prove no new publication.

Use only public feed URLs you are permitted to process. URLs and matching terms are stored in the run input; do not include secret tokens or private information. Feed content remains the publisher's content; downstream use must respect its terms. The tool does not verify factual claims in that content.