Silent workflow check for n8n

Finds scheduled n8n workflows that stopped running.

That failure produces no failed execution, so an error workflow never fires for it. It happens after a restart that did not re-register the schedule triggers, or when someone switches a workflow off while debugging and forgets it. The execution list looks clean and nothing has run for days.

What it does

Reads your own instance through the n8n API. For every active workflow that starts itself (Schedule, Cron or Interval trigger) it looks up the last run and compares it with that workflow's own schedule. A workflow counts as silent once twice its interval plus five minutes has passed without a run: a 15-minute sync is flagged after 35 minutes, a daily report after two days, and there is no threshold to set per workflow.

Test runs from the editor don't count. Opening a stalled workflow and clicking Execute creates a fresh execution, but the schedule still isn't firing, so the check skips executions with mode: manual and looks at the newest of the last 20 that aren't.

Workflows started from outside (webhooks, forms, chats, app triggers) are not judged. Idle is normal for them, and a check that calls idle broken gets muted within a week.

Every morning → Configuration → Get workflows → One item per workflow → Get last run
  → Find silent workflows → Heartbeat URL set → Ping heartbeat

It always returns exactly one item:

{
  "silent": 1,
  "summary": "1 of 9 checked workflows missed their schedule: Nightly invoice sync (last run 74 h ago, expected every day). 1 could not be judged, see notCovered.",
  "findings": [
    {
      "workflow": "Nightly invoice sync",
      "workflowId": "aBcD1234",
      "expected": "every day",
      "lastRun": "2026-09-08T02:00:11.000Z",
      "hoursSilent": 74,
      "note": "missed at least two scheduled runs"
    }
  ],
  "notCovered": [
    {
      "workflow": "Month start export",
      "workflowId": "eFgH5678",
      "reason": "Its schedule has no fixed rhythm, certain days of the month for example, so there is no gap to measure against."
    }
  ],
  "judged": 9,
  "notScheduled": 4,
  "checkSuspect": false,
  "hint": ""
}

Put an IF on silent greater than 0 after Find silent workflows and send summary to wherever your alerts go.

Setup

1. In n8n: Settings → n8n API → Create an API key.

2. Import silent-workflow-check.json via Workflows → Import from File.

3. On Get workflows, create a Header Auth credential:

   • Name: X-N8N-API-KEY
   • Value: your API key

   Use the same credential on Get last run.

4. In Configuration, set baseUrl to your n8n address without a trailing slash.

5. Optional, see below: set heartbeatUrl.

6. Attach your alert node and activate the workflow.

Watching the watcher

This workflow lives inside the instance it watches. If that instance is down, mid-restart, or its schedule triggers never re-registered, this check does not run either, and its silence looks exactly like everything being fine.

The heartbeat closes that gap. Create a free check at a dead man's switch service such as healthchecks.io, give it a period of one day, and paste its ping URL into heartbeatUrl. The workflow pings it at the end of every run, whether or not it found anything. When the pings stop, the service tells you. The ping node continues on error, so a heartbeat service that is down never swallows a real finding.

When it refuses to answer

Some workflows cannot be judged from execution history. The check lists them under notCovered with the reason instead of guessing:

• schedules without a fixed rhythm, certain days of the month for example;
• workflows set not to save successful executions: n8n deletes their healthy runs, so they would always look silent;
• workflows changed recently that have not run since;
• rare schedules with no saved run, where n8n may already have pruned the last one. A pruned execution is not evidence that something did not happen.

If half or more of the checked workflows look silent at once, checkSuspect is true and hint says why: either they really all stopped, which is what a restart that did not re-register the triggers looks like, or the check cannot see what it needs, most often an instance running with EXECUTIONS_DATA_SAVE_ON_SUCCESS=none. Open one of them in n8n before acting on it.

Known limits

• It lives inside the instance it watches. The heartbeat tells you the checker stopped, not which of your workflows did.
• It judges schedules, not work. A workflow that runs on time and produces nothing looks healthy to it.
• It reads up to 250 active workflows, one API page. Past that, some are not checked.
• The code nodes avoid arrow functions, template literals and optional chaining on purpose: the firewall in front of the n8n template portal rejects uploads that contain them.

Credit

The heartbeat, the not-covered answers and the hit-rate warning came from posts by moneywithjjcom on the n8n community forum. Skipping editor test runs came from nightly_runs on r/n8n.

Six more failures that produce no error

Stalled schedules are one of seven. The others: downstream rate limits, credentials that expired, webhooks nobody calls any more, workflows switched off during debugging, execution history pruned by EXECUTIONS_DATA_MAX_AGE, and daylight saving shifting a schedule by an hour twice a year.

Written up with the one check that catches each, at duskwatch.me/checklist.

Built by the people behind Duskwatch, which runs these checks from outside the instance across every client instance you manage. This workflow is the single-instance version, free and MIT, and there is nothing in it that phones home. Read the JSON.

Tested against n8n 2.37.