Job Feed Delta

Track new vacancies, field-level updates and confirmed closures across Greenhouse, Lever and Ashby in one JSON report. Monitor up to 25 boards at the launch price of $0.005 per report, plus the run-start event. Route hiring signals into recruiting workflows, job-board sync and n8n/Make automations.

Stable source IDs, before/after values and reusable snapshots keep downstream processing focused on what changed. Built-in recovery preserves the last healthy board state during temporary source failures.

Who it is for

Job-board operators keeping listings current, recruiters watching target companies and teams building programmatic hiring signals. Supply the boards you want to monitor and receive ready-to-route change events.

Input

{
  "boards": [
    {"provider":"greenhouse","slug":"cloudflare"},
    {"provider":"ashby","slug":"notion"},
    {"provider":"lever","slug":"spotify"}
  ]
}

The first call returns BASELINE events. Store nextSnapshot in your workflow, then pass it as previousSnapshot on the next call. This requires no manual work once the workflow is wired up. Snapshots belong to the caller; there is no shared server-side customer state and no cross-customer cache.

For n8n or Make: schedule → load your last snapshot → run Actor → fetch OUTPUT → route events → atomically save nextSnapshot. Never run two writers against the same snapshot concurrently. Use a single serial schedule per feed and queue overlapping requests. A platform schedule alone, with unchanged input, will repeatedly return baselines; it must be paired with a stateful workflow.

What the events mean

• BASELINE: already present when observation started.
• NEW: new source ID since the previous snapshot.
• UPDATED: a tracked field changed; previous and current values are included.
• CLOSED: a source vacancy is absent from two successful board observations.

Remote status and salary retain the source's structured values; unavailable values use null. Source IDs preserve distinct vacancies with matching titles. Each ATS keeps its own IDs, including during provider migrations.

API coverage and run settings

1–25 boards per call; 10,000 postings per board; 4 MB per upstream response; sequential board requests; three attempts for transient failures; 12-second request timeout; 150-second collection deadline for the entire feed. Unchecked boards receive RUN_TIME_BUDGET warnings and retain their previous state. Only fixed official ATS hosts are fetched. Arbitrary URLs, redirects and application POST requests are unsupported.

If more than 70% of a known board with at least three roles disappears, MASS_REMOVAL_GUARD preserves the entire prior board. Independently verify the change before enabling acceptMassRemoval. Even with that option, closure requires two successful observations. Boards not included in the current call remain in the supplied snapshot but are not newly checked. Large snapshots should be split into separate feeds.

The output focuses on vacancy fields and change events. Use public boards you are authorized to monitor and comply with source terms.

Output

The default key-value store's OUTPUT record contains the complete report, including events, warnings, board observations and nextSnapshot. The default dataset contains one summary item per report. It is not one row per vacancy. Fetch OUTPUT when you need the snapshot for the next check; exporting the dataset alone does not preserve workflow state.

METRICS contains execution measurements. A successful run can still include warnings for individual boards. Check these before interpreting a lack of changes as a lack of hiring activity.

Pricing

Launch price: $0.005 per report, covering up to 25 boards, plus $0.001 per Actor start event. The platform's start-event quantity depends on memory (one event per GB, minimum one); the default is 256 MB. Platform usage is included. There is no additional default dataset-item event charge. One dataset item is one report summary, not one vacancy.

Repeated successful checks are billable even if unchanged. A partial-success report is charged and identifies failed sources in warnings. If every source fails, no report event is charged; the start event can still apply.

Set a maximum run cost through Console or the API to control spending. The adapter uses the official SDK's report event and user spending limits.

Support: Evgeny Gertner. Include the run ID and affected provider/board, without applicant data or credentials.