RSS News Monitor — Deduplication & Change Tracking

Build one clean research list from several feeds, without repeatedly paying for the same unchanged headlines.

Read public RSS 2.0 and Atom 1.0 feeds, filter titles or categories by topic, deduplicate identical article URLs across feeds, and compare against a previous run. Every delivered item includes its source link, publication date when available, and source-feed provenance.

What buyers can build

• A daily research queue from industry blogs and official announcements.
• A newsletter editor's shortlist with repeated URLs removed.
• A company-news or release-announcement watchlist connected to an existing workflow.

The Actor delivers headline metadata, not rewritten news or full articles. It does not contact anyone, send alerts by itself or publish your newsletter.

Start with five headlines

{"feeds":["https://www.nasa.gov/feed/"],"maxItems":5}

Five delivered items cost $0.005, plus the displayed Actor start charge. No source API key, browser or proxy subscription is required. Supply direct feed URLs; a normal website homepage is not auto-discovered.

To filter a topic, add keywords such as ["launch","research"]. Matching is case-insensitive across titles and categories; it is a deterministic substring filter, not AI relevance scoring. Optional publishedSince takes a UTC ISO date-time; entries with missing/invalid dates are excluded by that filter rather than guessed.

Deduplication and change tracking

Duplicate normalized article URLs within the run produce one row with sourceFeeds listing every matching source. URL fragments are removed; query strings remain significant. Different URLs describing the same story are not semantically merged. When duplicate URLs have different titles, the first supplied feed wins and a warning is included.

The first run writes SNAPSHOT. On later runs pass the previous run's key-value store ID in previousSnapshotStoreId, keep feed URLs and filters identical, and enable changesOnly. Unchanged headlines are suppressed. Modified titles, dates or categories produce field-level before/after evidence in changes. The workflow must update that store ID each time; no global shared history or automatic scheduled alert is included.

New means not observed in the supplied snapshot, not necessarily published today. Feeds often expose only recent entries, so this is not a historical news archive. Capped baselines can make previously unseen entries appear new in a later run. Missing entries are not labeled deleted. For consistent comparisons, do not roll the date filter forward between runs.

Delivered fields and exports

Field	Meaning
title, url	Feed headline and original article link
feedTitle, sourceFeeds	Publisher/feed context and deduplication provenance
publishedAt, sourcePublishedRaw	Parsed UTC date and source date; null if unavailable
categories	Source categories, not inferred labels
changeType, changedFields, changes	New/changed/unchanged and field-level evidence
firstObservedAt, observedAt	Your workflow's observation times
warnings	Missing dates, conflicting titles and truncation notes

JSON and standard Apify dataset exports are available. OUTPUT-CSV is UTF-8 and spreadsheet-formula-safe; JSON retains original text. SUMMARY reports duplicate counts, skipped invalid entries, errors and coverage; DELIVERY records export completeness; DIAGNOSTICS contains unchecked feeds; SNAPSHOT supports your next comparison.

Pricing and bounds

$1 per 1,000 delivered feed items ($0.001 each) plus the displayed Actor start charge, with platform usage included and no monthly rental. Duplicate URLs, filtered-out items and suppressed unchanged rows are not charged as feed items. A blocked, malformed or inaccessible feed produces an uncharged diagnostic. The start charge can apply even if no item matches.

Up to 20 feeds, 1,000 delivered items per run (50 default), 1.5 MB per feed response and an approximately 90-second request budget. XML depth and element counts are bounded; DTDs and custom entities are rejected. Set Apify's maximum charge before starting. If the result or time limit intervenes, SUMMARY marks partial coverage. Snapshots retain up to 10,000 recent records.

Access and content boundaries

Uses public HTTP with a truthful Tender Delta user agent and respects robots.txt and rate limits. No login, anti-bot bypass, JavaScript rendering, feed discovery, pagination, RDF/RSS 1.0 or JSON Feed support. Robots failures are handled conservatively. Private-network targets, IPv6-only destinations and nonstandard ports are unsupported. Article pages are never fetched; no article body, email list, translation or AI summary is delivered. Only titles up to 1,000 characters and other feed metadata are retained.

Publishers' feed and content terms apply; a public feed is not a blanket license to republish articles. Source links and attribution are preserved so your team can review originals.

RSS 2.0 specification, Atom specification. Independent tool by Tender Delta, not affiliated with the feed publishers. Report reproducible issues in this Actor's Issues tab using input and a redacted run link.