GreenLight

In n8n, a step can fail inside a run that still reports success. A production workflow did exactly that for three weeks. Its failing step had “continue on fail” enabled, so the error traveled on as ordinary data, every run finished green, and nothing was sent.

The only clue was the clock (the run time had drifted from 1.2 to 10.5 seconds), and no dashboard showed it.

GreenLight reads an n8n instance through its API and compares each workflow with its own history, so there are no thresholds to configure. It runs four checks, for a step that failed inside a successful run, a workflow much slower than usual, a scheduled workflow that stopped, and one running far less often than before.

It refuses to judge without enough history, because a confident alert that's wrong costs more trust than a missed one. The first version guessed which workflows were scheduled from how evenly their runs were spaced; it passed every synthetic test and failed on a real instance, where a burst of manual tests looked like a schedule. Now it reads the trigger node.

Open source under the MIT License, at version 1.0.0, with more than 200 tests and CodeQL in CI. It works from the command line and as a local dashboard, in English and Spanish, and a public demo shows the real dashboard with invented data. So far it runs on one real instance, my own.

It can only compare the execution history the instance still keeps, so a short retention window limits it. Workflows that have never run are skipped. Detail inspection is sampled, and on a large instance a scan takes as long as the API takes to answer.