Comparison
Midwatch vs rolling your own
The standard advice is real: add an Error Trigger workflow, point it at Slack, maybe ping Uptime Kuma. Do it - it is free and it catches hard failures. This page is about the four failure classes that advice structurally cannot catch, verified against n8n's own docs.
Four failure classes DIY cannot catch
-
1. The workflow that stops firing. Nothing runs, so nothing errors, so the Error Trigger stays silent. You need something OUTSIDE the schedule watching for absence.
-
2. The run that succeeds but does nothing. A token expires, the run goes green, zero rows written. No error was thrown, so no error can be caught.
-
3. The instance going down. The Error Trigger runs inside n8n - it dies with the instance it is supposed to report on.
-
4. Config drift. Someone edits a node mid-month. Behavior changes, nothing errors, nobody is told.
And n8n Insights?
n8n's built-in Insights dashboard (Pro plans and up) shows executions, failure rates, run time, and time saved - for one instance, looking backward. As of July 2026 it sends no alerts of any kind, and it cannot see across the multiple instances an agency runs. It is a rearview mirror, not a watchkeeper. Useful - and not monitoring.
| Capability | Midwatch | Roll your own |
|---|---|---|
| Hard failures | Caught, classified by likely cause | Caught by an Error Trigger workflow - genuinely fine |
| Silent stops | Learned cadence detection | Not catchable from inside the instance |
| Zombie runs | Sampled item-count detection | DIY IF-node validation per workflow, per output shape |
| Instance down | Two-way dead-man switch | Uptime Kuma pings the URL - but nothing watches Kuma, and it cannot see workflow-level health |
| Drift | Hash-compared baselines with diffs | Git-based workflow backups catch it after the fact, if you diff them |
| Client reports | Branded ROI reports, shipped | Screenshots and a Looker Studio dashboard you maintain forever |
| Cost | $197/mo design-partner access | Free, plus your hours - forever |
If you run three workflows for yourself, roll your own - honestly. The math flips when other people's revenue runs through your fleet: every DIY layer is one more thing only you know how to maintain, and the failure modes it misses are exactly the ones that cost clients money. That is the gap Midwatch sells: absence detection, a watcher outside the instance, and a report your client can read.
Comparison verified against public pages on 2026-07-05 - tell us if something is out of date: hello@midwatch.ai
Past the point of DIY?
Become a design partner - 7 days free after a qualifying call, then $197/mo prepaid.