Career Site Job Listings Scraper: Greenhouse, Lever & More

Scrape job listings from any company career site built on Greenhouse, Lever, Ashby or Workday — four systems that between them power most company career pages. Paste a careers URL and get its open jobs back as clean data: every posting in one identical shape: job title, company, department, location, ISO country code, remote and workplace flags, employment type, publication date, the full job description as HTML and as plain text, the published salary range where the employer discloses one, and a direct apply link. You can mix boards from all four systems in a single run; the Actor works out which system each URL belongs to on its own.

Who it's for

• Job-market analytics — track hiring volume, locations, titles and disclosed pay over time.
• Recruiting and talent tools — keep a live picture of who is hiring for what.
• Job aggregators and job boards — fill your listings from the source, in one schema.
• Sales and talent intelligence — open roles are one of the clearest signals of what a company is building, funding and prioritising.

What it does

• Four job-board systems, detected automatically. Greenhouse, Lever, Ashby and Workday. Paste any mix of them; no per-site configuration.
• One schema for everything. 21 fields, same names and same types no matter where a job came from, so you can sort, filter and join across employers without cleaning anything up.
• Structured salary data. Where an employer publishes pay, you get min, max, currency and the pay interval as numbers — not a sentence to parse. The original text is kept too.
• Countries normalized to ISO codes. Boards write the same country five different ways ("US", "USA", "United States", "United States of America"). You get US in country and the board's own wording in countryName.
• Gets past Workday's 2,000-job ceiling. Workday never returns more than 2,000 jobs for one search. Switch on the deep crawl and the Actor searches category by category and merges the results — 2,615 jobs from NVIDIA's board in testing, against 2,000 without it.
• Remote and workplace flags. isRemote and workplaceType (on-site / hybrid / remote) come straight from the board where it publishes them.
• A bad URL doesn't sink the run. Each board is reported separately; the rest keep going.

Input

Only one field is required: the list of career-board URLs. Copy them from the employer's careers page — they look like boards.greenhouse.io/stripe, jobs.lever.co/spotify, jobs.ashbyhq.com/ramp or nvidia.wd5.myworkdayjobs.com/NVIDIAExternalCareerSite.

A realistic run — four fintech boards, salary data on, capped at 200 jobs each while you check the output:

{
    "startUrls": [
        "https://jobs.ashbyhq.com/ramp",
        "https://boards.greenhouse.io/stripe",
        "https://boards.greenhouse.io/coinbase",
        "https://boards.greenhouse.io/robinhood"
    ],
    "includeDescription": true,
    "includeCompensation": true,
    "maxJobsPerBoard": 200
}

Setting	What it does
Career board URLs	The boards to scrape. Mix systems freely.
Include job description	Full description as HTML and plain text. On by default. See the Workday note below.
Include compensation	Adds the salary object when the employer publishes pay. On by default, costs nothing extra.
Max jobs per board	Caps each board — the cheapest way to test before a full run. Starts at 25; 0 means everything.
Stop after this many seconds	Optional early stop. 0 uses the run timeout — the Actor always finishes itself just before it.
Fetch Greenhouse pay ranges	Off by default. Greenhouse hides pay behind a per-job page, so this costs one extra request per job. Only switch it on if you need it.
Workday deep crawl	Off by default. Only for Workday employers with more than 2,000 open jobs.
Include raw payload	Attaches the untouched board response to each record. Makes the dataset much larger.
Proxy configuration	Leave as it is. Datacenter proxy is included in your plan and is all these boards need.

Workday and descriptions. Workday is the only one of the four that serves descriptions one job at a time. With "Include job description" on, a Workday board costs one extra request per job and you get descriptions, publication dates and countries. With it off, those three fields come back empty for Workday jobs and the board is far faster and cheaper. The other three systems return descriptions with the listing, so the setting costs them nothing.

Long board lists and the run timeout. Boards are crawled in the order you list them, and the Actor stops itself shortly before your run timeout rather than being killed by it. When that happens the run still finishes normally: everything scraped so far is in the dataset, the log says which boards were not started, and the run status message says it stopped early. As a rough guide, a 2,000-job Workday board with descriptions on takes about 15 minutes, while Greenhouse, Lever and Ashby boards of any size take well under a minute each. If you are scraping many big boards, split them across several runs, cap "Max jobs per board", turn "Include job description" off, or raise the run timeout in the run options.

What you get back

One record per job. This is a real record from Ramp's board, with the description and the original pay payload shortened for readability:

{
    "source": "ashby",
    "company": "Ramp",
    "boardUrl": "https://jobs.ashbyhq.com/ramp",
    "jobId": "34413f8d-26bf-4bbc-8ade-eb309a0e2245",
    "title": "Security Engineer, Cloud",
    "department": "Engineering",
    "team": "Backend",
    "location": "New York, NY (HQ)",
    "secondaryLocations": ["Remote (Canada)", "Remote (US)", "Miami, FL"],
    "country": "US",
    "countryName": "USA",
    "isRemote": true,
    "workplaceType": "hybrid",
    "employmentType": "FullTime",
    "publishedAt": "2026-04-07T17:12:35.753Z",
    "updatedAt": null,
    "descriptionHtml": "<h1><strong>About Ramp</strong></h1><p>Ramp is building the smart infrastructure for finance teams…</p>",
    "descriptionText": "ABOUT RAMP\n\nRamp is building the smart infrastructure for finance teams…",
    "compensation": {
        "min": 211400,
        "max": 290600,
        "currency": "USD",
        "interval": "yearly",
        "raw": {
            "compensationTierSummary": "$211.4K – $290.6K • Offers Equity",
            "scrapeableCompensationSalarySummary": "$211.4K - $290.6K"
        }
    },
    "applyUrl": "https://jobs.ashbyhq.com/ramp/34413f8d-26bf-4bbc-8ade-eb309a0e2245/application",
    "jobUrl": "https://jobs.ashbyhq.com/ramp/34413f8d-26bf-4bbc-8ade-eb309a0e2245"
}

jobId is the employer's own identifier and stays the same between runs, so it works as a deduplication key. Download results as JSON, CSV, Excel or XML, or pull them from the Apify API.

What each system provides

Employers can only give you what their job board publishes. Here is what to expect, measured on live boards in August 2026.

Field	Greenhouse	Lever	Ashby	Workday
Title, location, apply link	yes	yes	yes	yes
Department	yes	yes	yes	—
Team	—	yes	yes	—
Country code	best-effort, from location text	yes (native)	yes (native)	yes (needs descriptions on)
Remote / workplace type	guessed from location wording	yes (native)	yes (native)	guessed from location wording
Employment type	—	yes	yes	yes
Published date	yes	yes	yes	needs descriptions on
Last-updated date	yes	—	—	—
Description	yes	yes	yes	needs descriptions on
Salary range	rare, and only with the pay option	rare	usually	never published

Salary coverage we actually measured. Ashby is far ahead: 117 of Ramp's 122 jobs carried a structured pay range. Lever publishes pay on a small minority of postings (1 of 25 sampled at Spotify). Greenhouse is uneven — GitLab published a range on 13 of 40 jobs sampled, Stripe on none of those sampled, and it costs an extra request per job to find out. Workday exposes no salary data at all through its job board.

Countries. Lever, Ashby and Workday all state the country, so those come back complete — we saw 25 of 25 populated for each in a mixed test run. Greenhouse states no country anywhere, so it is inferred from the location text and only when that text names a country outright: "Dublin, Ireland" becomes IE, but "San Francisco, CA" is left empty rather than guessed as Canada. On a Stripe sample, 5 of 25 jobs got a country this way. countryName always carries something readable when a country is known.

What a board costs to fetch. Roughly, per board:

System	Listing requests	Extra requests
Greenhouse	1 for the whole board (550 jobs at Stripe in one request)	+1 per job, only if you turn on Greenhouse pay ranges
Lever	1 per 100 jobs	none
Ashby	1 for the whole board (122 jobs at Ramp)	none
Workday	1 per 20 jobs — 2,000 jobs is about 100 requests	+1 per job for descriptions; deep crawl takes a 2,000-job board to roughly 150+ listing requests

Workday is the slow one by design: it caps pages at 20 jobs and refuses to return more than 2,000 results per search. Everything else is cheap.

Pricing

Pay per job returned. You are charged for each job record the Actor delivers — no job, no charge — plus your normal Apify platform usage. Capping "Max jobs per board" caps what you spend, and if a run hits your budget limit it stops cleanly and keeps the jobs it already collected rather than failing.

$0.002 per job listing. Pay-per-event pricing — you are only charged for the job rows the Actor actually delivers.

A note for Apify free-plan users: your plan comes with a monthly usage allowance, and a paid Actor draws on it like anything else. Start with "Max jobs per board" set to something small so you can see the output shape before committing a large run.

Fair use and how it behaves

• Public data only. Everything here comes from the same public job-board endpoints that render the employer's own careers page. No logins, no cookies, no accounts, no CAPTCHA solving, nothing behind a paywall.
• robots.txt is respected. Workday serves a different robots.txt for every employer, so it is fetched and honoured per employer at run time. A board that disallows crawling is skipped, and the run summary says so by name — it is never quietly reported as an employer with no jobs.
• Polite by default. Requests are rate-limited per host, crawl delays declared in robots.txt are obeyed, and failures back off instead of hammering.
• Honest identification. The Actor identifies itself in its user agent rather than pretending to be a browser.
• No personal data is collected beyond what employers publish about their own vacancies.

Limitations worth knowing

• Greenhouse salary data is sparse and expensive. It lives on each job's own page, so collecting it means one request per job, and plenty of boards publish none at all.
• Greenhouse publishes no country field. It is inferred from the location text, and only when that text actually names a country — so expect empty countries on boards that write "London" or "San Francisco, CA".
• Ashby's country is free text, so it is normalized on a best-effort basis. The ISO code is what we matched; countryName keeps the board's original wording so you can always check.
• Workday publishes no salary data, and its descriptions, publication dates and countries only arrive if you leave "Include job description" on.
• Workday caps searches at 2,000 results. The deep crawl is the way past it; without it, very large employers are truncated and the log says so.
• Boards move. Companies switch systems and retire board URLs; a URL that no longer exists is reported as a clear per-board error and the remaining boards still run.
• Only these four systems are supported today. Career pages built on anything else are rejected with an explanatory message.

Support

Found a board that behaves oddly, or want another job-board system supported? Open an issue on the Actor page with the board URL — that is the fastest way to get it looked at.

---

For developers

npm install
npm run build
npm test                    # unit tests, no network
apify run                   # reads storage/key_value_stores/default/INPUT.json

SPEC.md documents the verified behaviour of all four job-board APIs and the publishing checklist. Source layout: src/main.ts (run lifecycle and charging), src/detect.ts (URL → board), src/adapters/* (one per system), src/normalize.ts and src/countries.ts (field cleanup), src/http.ts (throttled client), src/robots.ts (robots.txt).