How do I know if my automation didn't run?

Short answer

Have each run check in when it finishes, tell an outside monitor when to expect that check-in, and get alerted when it doesn't arrive (or arrives with nothing in it). This is called a heartbeat, or a dead man's switch. Did It Run? watches your scheduled agents and automations this way and alerts you when a run is missed, late, failed, or comes back empty.

Why you didn't get an alert

Most platforms alert on *errors*: a run started, a step threw, you get an email. The failures that hurt are the ones where nothing starts, so there's nothing to report:

• The trigger stopped firing. An expired OAuth token, a stale webhook, an IMAP trigger that dropped its connection. Often no execution is created, so the platform's error alert never fires.
• The workflow, Zap or scenario was switched off. A teammate paused it, the platform disabled it after repeated errors, or you ran out of credits or operations. No alert.
• The machine wasn't there. A laptop asleep at 2am, a VPS that rebooted, a container that was OOM-killed. The thing that would alert is down too.
• The schedule is wrong. 5 * * * * (hourly at :05) instead of */5 * * * *, a server clock in UTC, a daylight-saving jump that skipped 2:30am. It's doing exactly what you told it.
• It ran and produced nothing. A search returned 0 rows, an agent replied with a refusal, an empty catch {} swallowed the error. It reports success.

The fix: watch the outcome, not the error

1. Write down the expectation. "Every weekday by 9:15am New York time", "every 5 minutes", or "at least once a day" for event-driven workflows. Add a grace window for normal delays (GitHub Actions schedules are often late; 30–60 minutes is sensible there).
2. Check in at the end of every successful run. One HTTP request to a unique ping URL as the last step. Optionally ping /start at the beginning and /fail from your error path, so failures alert immediately instead of at the deadline.
3. Send a little of the output. An item count (?items=12) or the result text. Then "ran but returned nothing" becomes an alert instead of a green checkmark.
4. Keep the monitor outside the thing it watches. A "did it run?" check inside the same n8n instance or the same server dies with it.
5. Get one alert, then a recovery notice. Not an alert every minute.

Copy-paste check-ins

cron or any script:

15 9 * * 1-5  /opt/jobs/digest.sh && curl -fsS -m 10 --retry 3 https://diditrun.app/p/<your-token> > /dev/null

Report an item count (catches empty runs):

count=$(wc -l < /var/data/today.csv)
curl -fsS -m 10 --retry 3 "https://diditrun.app/p/<your-token>?items=$count"

• n8n: add an HTTP Request node at the end of the workflow (POST to the ping URL). n8n's Error Trigger only fires when an execution fails; it can't see a trigger that stopped creating executions. See the n8n recipe.
• Zapier: add Webhooks by Zapier → POST as the last step. Webhooks needs a paid Zapier plan (Zapier's Free plan doesn't include it). See the Zapier recipe.
• Make: finish with HTTP → Make a request and add an error-handler route that calls /fail. See the Make recipe.
• GitHub Actions: curl the ping URL in a final step with if: success(), and /fail with if: failure(). Store the URL as a secret and create the watch in UTC. See the GitHub Actions recipe.
• ChatGPT and Claude scheduled tasks: where the agent can run a shell command, have it hit the ping URL as its last action. ChatGPT tasks can't make the request themselves, so relay their notification email to the ping URL with an automation tool. See the ChatGPT and Claude Code recipes. Email check-ins are coming soon; they aren't live yet.

Event-driven workflows: "quiet day" vs "dead trigger"

A webhook-triggered workflow can legitimately do nothing on a slow day. Two options:

• Use the longest normal gap as the schedule ("at least once every 24 hours"). You'll hear about a dead trigger within a day.
• For critical flows, send a synthetic test event on a schedule (a test form submission, a test record with a marker) and check that the workflow processed it.

Doing it with Did It Run?

1. Create a free account and add a watch: pick the schedule in plain English or as a cron expression, a timezone, and a grace window.
2. Paste the ping URL into the last step of your job (recipes for n8n, Zapier, Make, GitHub Actions and cron).
3. Optionally set a minimum item count, minimum size, or required keyword so empty runs raise an empty alert.
4. Get alerts by email (Free) or Slack, Discord and Telegram (Solo and Pro), plus a daily roll-call of what ran, what's missing and what came back empty (Solo and Pro).

Prefer a developer-first tool or self-hosting? Healthchecks.io, Cronitor, Dead Man's Snitch and UptimeRobot's heartbeat monitors all do the core check-in pattern well; see our heartbeat monitor comparison.