ats-boards

Discover and read the public job boards of the applicant-tracking systems German employers use — without scraping job portals.

Most companies do not host their own job list. The career page embeds a board on Personio, SmartRecruiters, Greenhouse, Workday, Oracle, softgarden … and those boards expose public, keyless endpoints. This package finds the board link in a career page and reads the postings from it, in one common shape.

Provider	Discovery pattern	Read from	Verified against
Personio	{code}.jobs.personio.de, jobs.personio.de/{code}	/xml feed	clark
SmartRecruiters	careers.smartrecruiters.com/{Code}	posting API	BoschGroup (790 DE postings)
Greenhouse	boards.greenhouse.io/{code}, embed ?for=	boards API	celonis
Lever	jobs.lever.co/{code}	postings API	kolibrigames
Ashby	jobs.ashbyhq.com/{code}	posting API	statista, parloa
Recruitee	{code}.recruitee.com	offers API	everdrop
Teamtailor	{code}.teamtailor.com	JSON feed	—
JOIN	join.com/companies/{code}	public JSON	—
Workable	apply.workable.com/{code}	widget API	—
Workday	{tenant}.wd{N}.myworkdayjobs.com/{Site}	CXS JSON (Germany facet)	db.wd3/DBWebsite (257)
Oracle Recruiting	{host}.fa.{region}.oraclecloud.com/…/sites/CX_…	recruitingCEJobRequisitions	Sana Kliniken (1,117)
softgarden	{code}.softgarden.io	JobPosting JSON-LD on posting pages	visiolife-group

Only Germany-located postings are returned; the location filter is intentionally conservative (see germany.ts).

Install

npm i ats-boards

Node 22+. No runtime dependencies.

Try it without writing code

npx ats-boards https://www.scalable.capital/en/careers
# Found 1 board(s): smartrecruiters/ScalableGmbH
# smartrecruiters/ScalableGmbH: 94 posting(s) in Germany
# (Associate) Product Manager (m/f/x)   Berlin   2026-09-03  https://jobs.smartrecruiters.com/ScalableGmbH/744000147296099
# ...

npx ats-boards greenhouse celonis          # read one board directly
npx ats-boards workday db.wd3/DBWebsite --json > db.json

Discovery works on server-rendered career pages that link to the board. If a page loads its jobs with JavaScript, pass the provider and board code directly.

Use

The recommended API uses English field names. The original API remains available unchanged for existing users.

import { discoverJobBoards, readJobBoard, setUserAgent } from 'ats-boards'

// Identify yourself to the sites you read.
setUserAgent('Mozilla/5.0 (compatible; MyCrawler/1.0; +https://example.com/bot)')

// 1. Find boards in a career page you already fetched.
const html = await (await fetch('https://www.example-gmbh.de/karriere')).text()
const boards = discoverJobBoards(html)
// → [{ provider: 'personio', code: 'example-gmbh', url: 'https://example-gmbh.jobs.personio.de/' }]

// 2. Read the postings.
for (const b of boards) {
  const result = await readJobBoard(b.provider, b.code, {
    employer: 'Example GmbH',
    companyDomain: 'example-gmbh.de',
  })
  if (result.ok) {
    for (const p of result.postings) console.log(p.title, p.locations[0]?.city, p.url)
  } else {
    console.warn(result.reason, result.message)
    // 'access' | 'permission' | 'rate_limit' | 'format'
  }
}

readJobBoard returns { ok: true, postings: [] } for a board that no longer exists (HTTP 404) — that means “no postings”, not a transport error. Rate limits return { ok: false, reason: 'rate_limit' }.

Posting shape

type JobPosting = {
  provider: string
  providerPostingId: string // "{boardCode}:{jobId}", stable per provider
  title: string
  employer?: string
  companyDomain?: string
  locations: { city?: string; postalCode?: string; country: string }[]
  occupation?: string
  description?: string      // HTML stripped when the provider supplies it
  employmentType: 'full_time' | 'part_time' | 'shift' | 'unknown'
  publishedDate?: string    // YYYY-MM-DD
  url?: string
  annualGrossEur?: number
}

Existing API

discoverBoards and readBoard keep their original Turkish result fields for backward compatibility. New integrations should prefer discoverJobBoards and readJobBoard; toJobPosting converts an existing Posting value without another network request.

Notes on behaviour

• Every request has a 15 s timeout and a descriptive user agent.
• List endpoints are paged up to 10 pages (20 for Workday). Workday and Oracle need one extra request per posting for the full text; those are read in a rotating window of 40 per call so a large board is covered over a day rather than in one burst.
• Only what the provider returns is used. Nothing is inferred; if a board does not give a description, that field stays empty.
• These are the same endpoints the providers' own career pages call. Check each provider's terms for your use case; this package does not bypass any authentication.

Examples

• examples/career-page.mjs — career page URL in, postings out
• examples/to-csv.mjs — one board to CSV
• examples/many-employers.mjs — several boards, keep postings with full text

Development

npm test            # offline discovery, contract and date tests
npm run build       # TypeScript → dist/
npm run check       # build + complete test suite

Adding a provider: a discovery rule in src/discover.ts, a reader in src/readers.ts (verify it against a real employer and note which one), and a test line in tests/discover.test.ts.

Contributions are welcome. See CONTRIBUTING.md for the provider checklist and SECURITY.md for private vulnerability reporting.

Why this exists

Built for Kelyvaro, a tool that helps job seekers from Turkey apply to German employers directly. The matching and application parts stay in the product; the board readers are generic and useful to anyone collecting German job postings, so they live here.

MIT © Murat Sert