Register expected jobs over HTTP
A receipt tells Crontinel a job ran. Registering the job tells Crontinel it should run. With both, you get an alert when a run never starts, even if nothing was posted.
No Laravel package is needed. Use the same app ingest key as the check-in recipes.
Register a job
curl -sS -X POST "https://app.crontinel.com/api/v1/ingest/schedule" \ -H "Authorization: Bearer $CRONTINEL_INGEST_KEY" \ -H "Content-Type: application/json" \ -d '{ "source": "named_by_user", "tasks": [{ "command": "nightly-import", "job_name": "nightly-import", "expression": "0 2 * * *", "timezone": "UTC", "grace_seconds": 300, "max_runtime_seconds": 3600, "counted_minimum": 1 }] }'| Field | Required | Meaning |
|---|---|---|
source | yes | named_by_user for HTTP runtimes. |
tasks[].command | yes | The exact command your receipts send. Runs are matched on it. |
tasks[].job_name | no | Stable name that survives a command change. If you set it, send the same job_name on every run receipt, or the job’s rules won’t apply to those runs. |
tasks[].expression | no | Five-field cron expression. Without it the job is inventory only and no missed-run check runs. |
tasks[].timezone | no | IANA time zone for the expression. Default UTC. |
tasks[].grace_seconds | no | How late a start may be before it alerts. Default 300, or 900 for a command starting with agent:. |
tasks[].max_runtime_seconds | no | How long a running receipt may stay open before it alerts as unfinished. Default 3600, or 7200 for a command starting with agent:. |
tasks[].counted_minimum | no | Creates a processed_records rule: fewer than this fails the run even on exit 0. |
If you register a job_name, add it to every receipt for that job, for example "job_name": "nightly-import" next to "command". Rules for a named job follow the name, so a receipt without it is treated as a different, unnamed job. Leave job_name out of both to match on command alone.
Scheduled AI agents get longer defaults because model calls vary. Name the job with an agent: prefix, such as agent:nightly-digest, and leave the two timing fields out. Values you send always win. See AI agent jobs for token, cost, and tool-call ceilings and the optional quality check.
The response lists created, unchanged, activated (an existing named job that gained its schedule), and skipped (plan limit). Posting the same body again is safe.
Node and Python
await fetch("https://app.crontinel.com/api/v1/ingest/schedule", { method: "POST", headers: { Authorization: `Bearer ${process.env.CRONTINEL_INGEST_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ source: "named_by_user", tasks: [{ command: "nightly-import", expression: "0 2 * * *", counted_minimum: 1 }], }),});import json, os, urllib.request
req = urllib.request.Request( "https://app.crontinel.com/api/v1/ingest/schedule", data=json.dumps({ "source": "named_by_user", "tasks": [{"command": "nightly-import", "expression": "0 2 * * *", "counted_minimum": 1}], }).encode(), headers={ "Authorization": f"Bearer {os.environ['CRONTINEL_INGEST_KEY']}", "Content-Type": "application/json", },)urllib.request.urlopen(req, timeout=5)What happens next
- Checking starts when you register. Earlier occurrences are not checked.
- If no receipt with that
commandstarts between the due time and the grace period, a cron never started alert opens. It resolves once the next scheduled run is observed. - If a
runningreceipt is still open aftermax_runtime_seconds, a cron never finished alert opens. See start and finish receipts. - If a run completes but took longer than
max_runtime_seconds, a cron ran too long alert opens. The next run inside the limit resolves it. check_monitoring_coverageover MCP names the registered jobs instead of reporting the inventory as unknown. It does not read crontab or an agent runner.
Laravel apps don’t need this request: php artisan crontinel:schedule (package v0.9.2 or later) registers the Laravel schedule the same way, with timezones. Earlier versions stored it as inventory only.