Website RSS & Atom Feed Finder

Resolve a public RSS website into validated RSS, Atom or JSON Feed endpoints. Supply website URLs and receive feed URLs, format, title, current entry count and per-input status. Use this endpoint inventory to set up ingestion pipelines—not to create feeds or download article bodies.

Who is it for?

Research teams assembling source lists, content operations configuring readers, and developers maintaining ingestion jobs can reuse the same website list on scheduled runs. Results identify working endpoints and distinguish inaccessible sites from tested sites without feeds.

Why use it?

A URL containing rss is not necessarily a feed. This Actor downloads each candidate and parses the document before accepting it. It follows redirects, resolves relative links, deduplicates final endpoints within each website, and exports source-attributed records. Empty but valid feeds are accepted with a zero entry count.

Getting started

1. Open the Actor input editor.
2. Add one or more public website or direct feed URLs.
3. Leave the discovery limits at their defaults, or lower them for a quick lookup.
4. Run and open the dataset.
5. Filter status=found to obtain usable feed URLs.
6. Review error/status rows before adding the sources to your reader.

Example input:

{
  "startUrls": [{"url": "https://www.nasa.gov/"}],
  "maxFeedsPerWebsite": 5,
  "maxCandidatesPerWebsite": 12,
  "probeCommonPaths": true
}

Input parameters

Field	Default	Behavior
startUrls	Required	1–100 public HTTP(S) website or direct feed URLs; URL strings or objects with url
maxFeedsPerWebsite	5	1–10 distinct validated final feed URLs per supplied entry
maxCandidatesPerWebsite	12	1–20 candidate fetches per website, excluding the initial page
probeCommonPaths	true	Try conventional paths after advertised links; shares the candidate limit

Only GET requests are supported. Request-list objects may contain url and method="GET"; custom headers, payload and userData are rejected even if your editor or hosted MCP offers those generic request-list options.

Duplicate supplied inputs are checked separately and can each incur a successful-website event. There is no unlimited mode. Direct feed input returns one feed without common-path probing. Query parameters are preserved when fetching public URLs; avoid supplying secrets.

How discovery works

First the initial URL is tested as a feed. Otherwise, an HTML document is inspected for alternate RSS/Atom/JSON links, HTTP Link headers, and anchors whose URLs resemble feed paths. Relative links honor a valid HTML base URL. Common-path discovery tries feed/, rss.xml, atom.xml, feed.xml, feed.json and rss/ at the page directory and site root.

Candidates are visited in that order. Final URLs after redirects are deduplicated within each input. A feed shared by different supplied websites remains attributed to each website. Discovery does not recursively crawl internal pages, render JavaScript, log in or solve bot challenges.

Output fields

Each validated feed produces one dataset row. If an input has no validated feed, it produces one uncharged status-only row with null feed fields.

Field	Meaning
inputUrl	Normalized supplied URL
siteUrl	Final initial-page URL, null if unavailable
feedUrl	Final validated endpoint, null for status-only rows
format	rss, atom, json, or null
title	Feed title, up to 500 characters; nullable
recordCount	Entries in this downloaded feed document; nullable
status	found, not_found, or error
httpStatus	Feed response status for found rows; initial-page status otherwise
discoveryMethod	direct, html_link, http_link, anchor, common_path, or null
candidatesChecked	Candidate URLs attempted for the input; direct feed counts as one
truncated	Some candidates were not tested because a cap was reached
error	Bounded error code, or null
checkedAt	UTC check-start timestamp
billingUnit	True on first validated feed row per input; false on included additional feeds and status-only rows

recordCount is not the site's total historical article count, subscriber count or count of new posts. No article records are exported.

Output example

A representative NASA check returned this RSS endpoint and ten entries. Entry counts can change as publishers update their feeds.

{
  "inputUrl": "https://www.nasa.gov/",
  "siteUrl": "https://www.nasa.gov/",
  "feedUrl": "https://www.nasa.gov/feed/",
  "format": "rss",
  "title": "NASA",
  "recordCount": 10,
  "status": "found",
  "httpStatus": 200,
  "discoveryMethod": "html_link",
  "candidatesChecked": 1,
  "truncated": true,
  "error": null,
  "billingUnit": true,
  "checkedAt": "2026-10-10T14:07:34.415Z"
}

This example used a one-feed cap; truncated=true means other candidates remained untested, not that the accepted feed was invalid.

How much does it cost to find feeds on websites?

Pay-per-event pricing consists of one start event per valid run and one item event per supplied website with at least one validated feed. Additional feeds from that same input have no separate event charge. No-feed/error status rows are free of item charges; the start charge still applies. Apify platform usage may be charged separately according to your plan.

The start event is $0.0001. Successful-website unit prices by Apify Store spend tier are FREE $0.0015502, BRONZE $0.001348, SILVER $0.0010514, and GOLD/PLATINUM/DIAMOND $0.0008088. Tiers depend on qualifying aggregate monthly Store spend, not volume in this Actor. At BRONZE, illustrative totals are 1 × $0.001348 + $0.0001, 10 × $0.001348 + $0.0001, or 100 × $0.001348 + $0.0001. These are Actor-event estimates, not a platform-usage estimate.

Set a maximum total charge in Apify to control event spending. At that limit the Actor stops collection rather than emitting uncharged useful output, so later inputs may be absent. Downloaded source entry counts do not affect event pricing.

Limits and errors

Each response is capped at 2 MiB. A request has at most five redirects, two attempts for transient network/429/5xx failures, a ten-second attempt timeout and a fifteen-second total request budget. Stable 4xx errors are not retried. Candidate scheduling stops after thirty seconds per website or 240 seconds per run; an already-running request may finish after that boundary. Remaining inputs receive runtime_budget status rows unless a billing cap stops the run first.

not_found only means no feed was validated in the tested scope; it does not prove the site publishes none. truncated flags untested candidates. error identifies unreachable pages, unsupported documents or incomplete candidate requests. If every supplied website errors, the run fails and retains its status rows. Mixed batches can succeed with errors; always inspect statuses.

Only anonymously reachable public websites are supported. Private/loopback/link-local addresses, embedded credentials and nonstandard ports are rejected, including redirect targets. DNS resolution is checked and pinned. DTDs and custom entity declarations are rejected; predefined and numeric XML entities in titles are safely decoded. Supported XML formats are RSS 2.0, RSS 1.0 with the RDF/RSS namespaces, and Atom with the Atom namespace. Non-UTF-8 or compressed responses that ignore normal uncompressed requests may not parse.

Integrations

Export JSON or CSV from the dataset to feed readers, spreadsheet source inventories or ingestion services. Use feedUrl as a stable lookup key and retain inputUrl for provenance. A scheduled repeat run can reveal changed endpoints by comparing inventories downstream; this Actor itself has no history, diffing or notification state.

In Make or n8n, run this Actor, filter found rows, then hand endpoints to your feed ingestion step. Keep error rows in a review queue rather than deleting those sources automatically. Multiple feed rows from one website represent one successful lookup for billing.

API use

Start a run and retain its run/storage IDs. Poll the same run until terminal, then read its dataset. Never restart merely because a client request timed out.

curl -X POST 'https://api.apify.com/v2/acts/automation-lab~website-rss-feed-finder/runs' \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H 'Content-Type: application/json' \
  -d '{"startUrls":[{"url":"https://www.nasa.gov/"}]}'

JavaScript with the Apify client:

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: process.env.APIFY_TOKEN });
const run = await client.actor('automation-lab/website-rss-feed-finder').call({
  startUrls: [{ url: 'https://www.nasa.gov/' }]
});
const page = await client.dataset(run.defaultDatasetId).listItems({ limit: 20 });
console.log(page.items);

Python:

import os
from apify_client import ApifyClient
client = ApifyClient(os.environ['APIFY_TOKEN'])
run = client.actor('automation-lab/website-rss-feed-finder').call(
    run_input={'startUrls': [{'url': 'https://www.nasa.gov/'}]})
print(client.dataset(run['defaultDatasetId']).list_items(limit=20).items)

MCP use

For authorized client use, select this stable Actor subset:

claude mcp add --transport http apify \
  'https://mcp.apify.com?tools=automation-lab/website-rss-feed-finder'

Equivalent Claude Desktop, Cursor and VS Code editor configuration:

{"mcpServers":{"apify":{"url":"https://mcp.apify.com?tools=automation-lab/website-rss-feed-finder"}}}

Configure Apify authentication in your client. Discover actual tool names and input schemas with scoped tools/list; Actor selection also exposes run/dataset/KV/abort helpers, not exactly one tool. Example prompt: “Find the public RSS, Atom or JSON feeds for these websites, show statuses and validated endpoint URLs.”

Start once and retain the run ID. call-actor waits at most 45 seconds; if pending, check that same run with bounded backoff (2, 4, 8 seconds, capped at 10) and a 120-second consumer deadline. Set the client timeout longer than the selected wait plus transport margin. At deadline, report pending and the run ID rather than starting again.

Report terminal failures honestly: any available dataset rows are partial results, not a successful full inventory. On success, call get-dataset-items for the same returned defaultDatasetId, initially with offset=0, limit=20, and the needed fields projection (for example inputUrl,feedUrl,format,title,recordCount,status,error,truncated,billingUnit), using the argument shape from tools/list. Read at most 100 source rows and 64 KiB of serialized UTF-8 content admitted to model context across all pages, whichever comes first. Enforce byte limits host-side before model injection; omit or summarize oversized rows with disclosure. If your client cannot intercept oversized replies, do not claim a hard byte guarantee. Track source offsets and report partial/truncated results and the continuation offset. Keep full downloads downstream, outside model context. These consumer limits do not change Actor collection limits.

Legality and privacy

Use only public URLs you are permitted to request. Respect publisher terms, applicable law and reasonable scheduling intervals. Public availability does not grant copyright or redistribution rights. Avoid placing tokens or private identities in URLs.

No AI model processes your input or source data in this Actor. There are no paid APIs or proxies. Results and logs use your Apify storage retention settings; delete the run and its storages in Console or via the API when no longer needed. There is no persistent source cache. For support, use the Actor's Store Issues tab.

Failed operations send sanitized diagnostic input, exceptions and Actor/build/run IDs to our private GlitchTip service for repair. Secret fields and URL queries are removed, and reports are retained for 30 days. Dataset output preserves supplied/public endpoint URLs for your use; treat your own exports appropriately.

FAQ and troubleshooting

Why does a website have no feed? It may not advertise one, its path may not be among the bounded guesses, or feeds may be on another page. Try a blog URL or a known feed URL and review truncated.

Does this create an RSS feed? No. It discovers and validates existing published feeds.

Can I read the articles? No. This product exports endpoint metadata and current entry counts. Use a feed reader downstream.

Why does a run fail even with dataset rows? If every website errors, the status rows are diagnostic results, not useful feeds. Fix URLs or inaccessible sources; do not treat the run as a successful full inventory.

Are empty feeds rejected? No. A valid empty feed has recordCount=0 and is usable as an endpoint.

Can I use proxy credentials or cookies? No. This version intentionally has no proxy, login or browser route.

Related Actors

For a source-specific article feed workflow, see Product Hunt Launch Feed Scraper. This Actor's endpoint inventory is complementary: discovery comes before choosing an appropriate reader or source-specific extractor.

This utility is not affiliated with or endorsed by source publishers or Apify.