Job Change Monitor: New and Closed Roles Only

Run it on a schedule. Get the four roles that appeared, not the 548 that were already there.

A search of the Apify Store on 2026-08-01 found no other Actor that reports job board changes. Everything else hands you the whole board and leaves the diffing to you.

{ "companies": ["stripe.com", "figma.com"] }

NEW     Staff Security Engineer          stripe.com
NEW     Product Designer, Payments       stripe.com
CLOSED  4318772008                       figma.com

No API key. No proxy. No diffing on your side.

Two things decide whether a monitor actually works

Both were checked, not assumed, and both are the reason this one can be trusted.

The memory has to be permanent

Apify's own documentation: "Named key-value stores are retained indefinitely. Unnamed key-value stores expire after 7 days."

A monitor built on the run's default storage silently loses its memory every week, and then reports an entire board as new. That is worse than no monitor, because it looks like it worked and it bills like it worked. This Actor keeps state in a named store, and the store name is an input so you can hold separate watchlists apart.

Two watchlists must never share a baseline

State is keyed per board, and the key carries a digest of the board token rather than only a sanitised version of it. Sanitising alone collapsed different boards onto one key: Bosch Group and Bosch/Group both became Bosch-Group.

Two boards on one baseline is not a small bug. Every run would report every posting of one board as new and every posting of the other as closed, then swap on the next run, forever, and you would pay for every one of those rows.

The first run must not bill you for the baseline

On day one everything is new by definition. Being charged for 548 "new" roles the first morning you point this at Stripe is a technicality dressed as a feature.

So the first run records the baseline, reports one row saying exactly that, and charges nothing:

Baseline recorded for stripe.com (greenhouse/stripe): 548 open roles.
This run is not charged. From the next run on, only roles that appear
or disappear are reported.

A role disappearing is a signal too

reportClosedRoles is on by default. A role leaving the board means it was filled, paused or withdrawn, and for anyone tracking a competitor's hiring that is often the more interesting half.

A failed check never corrupts the memory

If a board cannot be fetched, the remembered state is left alone and an error row is emitted. The alternative, overwriting the baseline with nothing, would make the next successful run report the whole board as new and charge for it. Verified: after a failed fetch the stored baseline is byte-identical.

Setting it up

1. Run it once. That records the baseline and costs nothing.
2. Add it to an Apify Schedule, daily or hourly.
3. Connect the dataset to Slack, email, a webhook or a sheet.

Changed which companies you watch? Set resetBaseline: true once.

What you get

New roles carry the full posting, in the same shape as ATS Jobs Scraper, plus changeType: "new". Closed roles carry changeType: "closed" with the job id, company, platform and board.

Field
changeType	new or closed
title, location, url	on new roles
departments, offices
company, atsPlatform, boardToken, boardUrl
postedAt, updatedAt
description	opt in, see below

Why this returns data when job scrapers return 403

Indeed, Glassdoor, ZipRecruiter and LinkedIn Jobs do not originate job data. They aggregate it from company career pages, and those pages run on systems that publish free, keyless, public APIs, because companies want their openings indexed. Measured from a residential address on 2026-08-01, indeed.com/jobs, glassdoor.com, ziprecruiter.com and upwork.com all answered 403 with zero bytes, while these sources answered with real jobs.

Going to the source is also faster to notice. An aggregator shows you its last crawl. This shows you the board.

Notes

• Covers Greenhouse, Lever, Ashby, Workable, Recruitee, Rippling, Workday, Personio and SmartRecruiters. Watch a company by domain on the first six; the last three take an exact board, because on those a wrong guess is indistinguishable from a right one.
• includeDescription is off by default because it is expensive at the source: Greenhouse returns 111 KB for a board without descriptions and 1.97 MB with them. Rippling's board API has no descriptions at any setting.
• Roles are matched by the ATS's own job id, so a retitled posting is not reported as a new one.
• Errors are per board. One unreachable company does not end the run, and error rows are not charged.
• Just want the current board: ATS Jobs Scraper. Just want to know which system a company runs: Hiring Tech Stack Detector.