Hiring Signal Radar — ATS Job Board Monitor + AI Tagging

Know the moment a target account starts hiring for a role that matters to your pitch.

Hiring Signal Radar tracks company job boards on Greenhouse, Lever, and Ashby — the three ATS platforms that expose public job APIs — and reports newly-opened postings since the last run. With an Anthropic API key set, each new posting gets a Claude-generated role category, seniority estimate, and a one-sentence buying-signal insight (e.g. "first DevOps hire suggests they're scaling infra now").

---

Why ATS APIs instead of scraping LinkedIn

Most hiring-signal tools scrape LinkedIn or Indeed, which means constant breakage against anti-bot defenses. Greenhouse, Lever, and Ashby all expose public, unauthenticated JSON APIs for their customers' job boards — no login wall, no bot detection, no proxy required. That makes this Actor cheap to run and low-maintenance, at the cost of only covering companies that use one of those three ATS platforms (a large and growing share of tech/startup hiring).

The output is clean structured JSON per posting — plug it straight into an AI agent, a CRM enrichment pipeline, or Zapier.

---

Quickstart

1. List companies to track

[
  { "name": "Notion" },
  { "name": "Ramp" },
  { "name": "Ro", "leverSlug": "ro" }
]

If you know the exact board slug, set greenhouseSlug / leverSlug / ashbySlug for a guaranteed match. Otherwise the Actor auto-detects the board by trying name-derived slug guesses against all three ATS platforms and caches whichever one resolves.

2. (Optional) Add keywords to watch

watchKeywords: e.g. ["DevOps", "Enterprise Sales", "Data Platform"] — matching postings get flagged with 🎯 in Slack. Leave empty to track everything.

3. (Optional) Add your Anthropic API key

Get one at console.anthropic.com. Without it, you still get every new posting — just without the AI role/seniority/buying-signal tags.

4. Run it once to capture a baseline

The first run per company just records the current open-jobs list — no "new posting" output yet, since there's nothing to diff against.

5. Set up a schedule

Apify Schedules tab → daily or every few hours. From then on, new postings are detected automatically.

---

Output

One dataset record per new posting:

{
  "company": "Ramp",
  "ats": "ashby",
  "boardSlug": "ramp",
  "job": {
    "externalId": "34413f8d-26bf-4bbc-8ade-eb309a...",
    "title": "Security Engineer, Cloud",
    "department": "Engineering",
    "location": "New York, NY (HQ)",
    "url": "https://jobs.ashbyhq.com/ramp/...",
    "postedAt": "2026-07-30T00:00:00.000Z"
  },
  "tag": {
    "roleCategory": "DevOps/Infra",
    "seniority": "Senior",
    "buyingSignalInsight": "A senior cloud security hire signals Ramp is investing in infra hardening — good timing for security tooling or compliance pitches."
  },
  "matchedKeywords": ["Security"],
  "detectedAt": "2026-07-31T14:00:00.000Z"
}

---

Pricing guidance

Billed per company tracked per run (pay-per-event recommended). Each company costs one or two lightweight JSON API calls plus, if AI tagging is enabled, one Claude Haiku call per run covering all that company's new postings in a single batch — not one call per job.

---

Known limitations

• Only covers Greenhouse, Lever, and Ashby. Companies on Workday, LinkedIn-only, or a custom careers page won't resolve — the Actor reports an error for that company rather than guessing.
• Auto-detect is name-based guessing. It tries the company name as a slug across all three platforms; if the company's actual slug is different (rebrand, abbreviation), set the explicit *Slug field.
• A company changing ATS platforms will look like every job is "new" on the next run relative to the old snapshot — this is a one-time false signal, not an ongoing issue.

---

Architecture

Greenhouse / Lever / Ashby public JSON APIs → normalized job list
  → Apify KV Store (previous job-ID snapshot per company)
  → Diff engine (new postings since last run)
  → Claude Haiku (batched role/seniority/buying-signal tagging)
  → Slack Block Kit digest
  → Apify Dataset (full run archive)