ATS Job Board Aggregator — Greenhouse + Ashby + Lever + Workday

Pull live job postings straight from the official public job-board APIs of the ATS platforms most companies use. Paste a company name, a board slug, or a full careers URL — get one normalized dataset back.

No login. No API keys. No scraping. No browser automation.

If you were about to write a job board scraper for Greenhouse, Ashby, Lever or Workday — this is the same data, read from the vendor's own JSON endpoint instead of the rendered page. Same fields, one input, and no HTML selectors to repair every time a careers page is redesigned.

Provider	Endpoint used	Verified (2026-10-11)
Greenhouse	boards-api.greenhouse.io/v1/boards/{slug}/jobs	stripe → 729 jobs, gitlab → 221
Ashby	api.ashbyhq.com/posting-api/job-board/{slug}	openai → 815, ramp → 165
Lever	api.lever.co/v0/postings/{slug}?mode=json	leverdemo → 10
Workday	{host}/wday/cxs/{tenant}/{site}/jobs (POST)	nvidia → 2,000

These are the same endpoints those vendors publish so companies can embed their job board on their own website. They return clean JSON, so this Actor does not break when a careers page is redesigned — a failure mode that kills HTML scrapers.

Workday matters if you need enterprise hiring data: it is the ATS behind a large share of the Fortune 500, and a single tenant can carry 2,000+ open roles.

Input

Field	Type	Default	Notes
companies	array	["stripe","openai","gitlab","ramp"]	Three accepted formats — see below
keywords	array	[]	Match against title / department / team / location
excludeKeywords	array	[]	Inverted filter
locationFilter	string	""	Case-insensitive substring, e.g. "Berlin"
remoteOnly	bool	false	Keep remote postings only
includeDescription	bool	false	Adds up to 20k chars of plain-text description
maxItems	int	2000	Hard cap
concurrency	int	4	Boards fetched in parallel
requestDelayMs	int	250	Polite pause between boards

The four ways to specify a company

stripe                                  → auto-detects which ATS it uses
ashby:openai                            → explicit provider
https://job-boards.greenhouse.io/gitlab → parsed straight from a careers URL
workday:nvidia.wd5.myworkdayjobs.com/NVIDIAExternalCareerSite
https://nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAExternalCareerSite

With auto-detect, the Actor tries Greenhouse → Ashby → Lever and keeps the first board that answers. A board that exists but currently has zero openings still counts as a hit — so you can tell "wrong slug" apart from "nothing open right now".

Workday is the one exception: it can't be auto-detected, because the tenant and site are defined by the customer and can't be derived from a company name. Give it either a workday:HOST/SITE target or the company's careers URL — just paste the URL from your browser and the Actor picks out the pieces.

Output

One uniform record per posting, whichever ATS it came from:

{
  "provider": "ashby",
  "company": "ramp",
  "companySlug": "ramp",
  "boardUrl": "https://jobs.ashbyhq.com/ramp",
  "jobId": "34413f8d-…",
  "title": "Security Engineer, Cloud",
  "department": "Engineering",
  "team": "Backend",
  "employmentType": "FullTime",
  "location": "New York, NY (HQ)",
  "allLocations": ["New York, NY (HQ)", "Remote (Canada)", "Remote (US)"],
  "isRemote": true,
  "url": "https://jobs.ashbyhq.com/ramp/…",
  "applyUrl": "https://jobs.ashbyhq.com/ramp/…",
  "postedAt": "2026-04-07T17:12:35.753Z",
  "updatedAt": null,
  "compensation": "USD 180000–240000 / year"
}

Results land in the default dataset; a SUMMARY key-value record reports which boards were found, how many jobs each returned, and which slugs failed.

Use cases

• Job boards & niche aggregators — build a vertical board from the companies you care about.
• Recruiting / sourcing — watch a competitor's hiring signals ("they just opened 12 infra roles").
• Sales & market signals — headcount growth by department, market by market.
• AI / research datasets — normalized, multi-company, multi-ATS job corpus with dates and locations.

Notes & limits

• No HR email addresses. These endpoints publish postings, not contact emails. Any Actor claiming otherwise is guessing.
• About includeDescription. Plain text, HTML stripped, capped at 20,000 characters per posting. Measured 2026-10-14 on live boards: Greenhouse stripe returns ~4.6k chars per posting, Ashby openai ~6.9k, Lever leverdemo ~1.4k. Two things worth knowing before you switch it on:
  • Ashby and Lever have no "no description" mode. Their public endpoints always return the text, so the toggle only decides whether it is written to your dataset. Only Greenhouse actually saves you bandwidth when it is off (stripe: 451 KB → 5.6 MB with content on).
  • Descriptions are the biggest cost lever, because pricing is per returned posting, not per byte — enabling it does not change what you pay, but it does change dataset size.
• A wrong board slug is a hard 404, not an empty result. Verified on Greenhouse: boards-api.greenhouse.io/v1/boards/<nope>/jobs → 404 {"status":404,"error":"Job not found"}. The Actor reports it under failures in the SUMMARY record, so "wrong slug" never gets confused with "this company is not hiring". A board that answers with zero postings is still a successful hit.
• department / team are populated for Ashby and Lever, but usually null for Greenhouse — Greenhouse's board endpoint stopped returning per-job department objects. Verified, not assumed: 950 Greenhouse postings in a test run came back without departments while the Ashby/Lever ones all carried them.
• Which ATS does a company use? Not provided by any API; auto-detect answers it by probing. Two calls, at most, for the common cases.
• Workday dates are approximate. Its list endpoint returns relative strings ("Posted Today", "Posted 30+ Days Ago"), so postedAt is normalised to that day's midnight UTC and the raw string is kept in postedText. Enable Include job description to replace it with the exact startDate from the detail endpoint.
• Workday locations can read "2 Locations". The list endpoint collapses multi-site postings; the full list is only in the detail endpoint. allLocations therefore holds one entry for Workday.
• Five other ATS platforms are deliberately not supported. Workable (the widget endpoint returned an empty jobs array for six real slugs), SmartRecruiters (keys on internal company IDs — returned totalFound: 0 for Visa, McDonald's and Ubisoft), Recruitee (401), BambooHR (302), Personio (XML only), Teamtailor (JSON Feed with no per-company endpoint). Listing a source that silently returns nothing is worse than not listing it — every provider here was verified against a live board first.
• Only public, documented endpoints. No authentication, no rate-limit circumvention, no anti-bot evasion.

Tested (2026-10-11, direct connection, strict TLS)

Case	Result
stripe (auto-detect)	greenhouse / 729 jobs
gitlab (auto-detect)	greenhouse / 221 jobs
openai (auto-detect)	ashby / 815 jobs
ramp (auto-detect)	ashby / 165 jobs
lever:leverdemo (explicit)	lever / 10 jobs
https://job-boards.greenhouse.io/gitlab (URL)	greenhouse / 221 jobs
workday:nvidia.wd5.myworkdayjobs.com/NVIDIAExternalCareerSite	workday / 2,000 jobs
https://nvidia.wd5.myworkdayjobs.com/en-US/NVIDIAExternalCareerSite (URL)	workday / 2,000 jobs
ashby:nonexistent-board-xyz	clean error: ashby:404
workday:this-does-not-exist.wd5.myworkdayjobs.com/NoSite	clean error: workday:422
zzz-not-a-real-company	clean error, all three auto-detect providers probed
Full local regression, 8 boards	8 ok / 0 empty / 3 expected errors — 61 seconds wall clock

Workday pagination, measured: the list endpoint caps at 20 postings per request and returns 0 rows (not an error) if you ask for more — a silent truncation trap. We page by 20 and fetch pages concurrently, which brought a 2,000-posting board from 3m49s to 68s.

Descriptions, measured separately (2026-10-14): the Greenhouse endpoint does not include the posting body by default — the bare response carries 14 fields and no content at all, and only ?content=true adds it (stripe: 451,481 B → 5,630,572 B). That parameter is now always sent when the toggle is on, and the escaped HTML (<h2>…) is decoded before tag-stripping, so what you get is prose, not markup. Regression suite: src/test-description.js, 27/27 assertions, covering the three providers plus the 404 semantics.

Pricing

Pay-per-event, in two parts:

Event	Price	What it covers
Actor start	$0.00005 per run	Booting the container (platform default — Apify still covers the first 5s of compute)
Result returned	$0.002 per posting	Each row pushed to the dataset

That is ~$2 per 1,000 postings, plus a nominal start fee. A 50-company run returning 1,000–3,000 postings costs $2–6.

For context: the most-installed Actor in this category charges $0.012 per posting — six times more for largely the same public endpoints. We're pricing for adoption first, and will revisit once there is usage data.