RSS Change Monitor

Watch RSS and Atom feeds and get only what changed since the last run. Schedule it, point a webhook at it, and stop re-processing the same items every time.

Charged per feed successfully checked. A feed that fails to fetch or isn't a feed is reported with the error and never charged.

---

Why not just parse the feed?

Parsing RSS is a library call. What you actually want scheduled is "tell me what's new" — and that needs state carried between runs, which no parser gives you. This Actor keeps a fingerprint of every feed in a named key-value store, so consecutive runs can diff against it.

That state is the whole product, and getting the diff right is harder than it looks.

The four things naive change-detection gets wrong

1. Item identity. Many feeds re-stamp <link> with utm_* tracking parameters on every fetch, and some republish items with a fresh <guid>. Diff on the link, or on a hash of the whole item, and you'll get change alerts for items that never changed. We resolve identity in order: guid → id → link with tracking parameters stripped → title + date.

2. New versus edited. A corrected article isn't a new one. We keep a separate content fingerprint per item, so edits come back as updatedItems and genuinely new items as newItems. You decide which matters.

3. Items disappearing is not deletion. Feeds carry only the latest N entries, so items scroll off the bottom constantly. Reporting those as "removed" would be a lie by construction, so we don't report removals at all.

4. The first run. There's no previous state, so nothing has changed. Tools that report every existing item as new on run one flood your webhook the moment you schedule them. Our first run records a baseline, reports zero changes, and says so explicitly in the output.

Both formats. RSS (<item>, <pubDate>, <link>) and Atom (<entry>, <updated>, <link href>) are both handled. A monitor that only reads RSS silently misses half the web.

---

Input

{
  "feedUrls": [
    "https://news.ycombinator.com/rss",
    "https://example.com/blog/atom.xml"
  ],
  "stateStoreName": "my-watchlist"
}

Field	Type	Default	Notes
feedUrls	array	—	Feeds to watch
feedUrl	string	—	Watch a single feed
stateStoreName	string	rss-monitor-state	Where the baseline lives. Different names keep independent watchlists apart.
resetBaseline	boolean	false	Forget history and start fresh. That run reports no changes.
maxNewItemsPerFeed	integer	50	Caps output size on busy feeds. Counts stay exact.
maxFeeds	integer	25	Cap on feeds, and therefore on spend

Output

One record per feed, plus a SUMMARY:

{
  "feedUrl": "https://example.com/feed.xml",
  "ok": true,
  "feedTitle": "Example Blog",
  "format": "rss",
  "status": "changed",
  "isFirstRun": false,
  "itemsInFeed": 30,
  "newCount": 2,
  "updatedCount": 1,
  "newItems": [
    {
      "id": "https://example.com/post-42",
      "title": "The post that just went up",
      "link": "https://example.com/post-42",
      "published": "Sun, 27 Jul 2026 18:04:00 GMT",
      "author": "Jane Doe",
      "summary": "First 600 characters of the description…"
    }
  ],
  "updatedItems": [],
  "checkedAt": "2026-07-27T20:31:00.000Z"
}

status is baseline on the first run for a feed, then changed or unchanged.

Scheduling

Set an Apify schedule (hourly, daily, whatever suits) and attach a webhook on run success. Every run after the first tells you exactly what appeared or changed. Keep the same stateStoreName across runs — that's what makes the diff possible.

For agents and automation

• Capability: detect new and updated items across RSS/Atom feeds between scheduled runs
• Required input: feedUrl or feedUrls
• Returns: one record per feed plus a SUMMARY; newItems / updatedItems are the payload
• Stateful by design: the named key-value store persists between runs. Same store name = same watchlist.
• Bounded: maxFeeds caps the run and the spend
• Side effects: reads feeds, writes fingerprints to your own key-value store. Nothing external.
• Failure: a bad feed is returned as a record with ok: false and an error; the run continues and that feed is not charged.

Pricing

Pay per feed successfully checked. Failed fetches are free.