Hey everyone,
I kept running into two problems with my production n8n workflows that the native dashboard doesn’t fully solve:
Problem 1 — AI cost is invisible
I had no idea which workflows were burning through my OpenAI/Anthropic budget until the bill arrived. Native execution logs show runs, not cost.
Problem 2 — Silent failures
A scheduled workflow stopped running. No error, no alert — just silently dead. I found out days later.
So I built a small tool that connects to your n8n instance via API and:
**•** Shows AI/API cost per workflow per run
**•** Alerts you (email or Slack) when a workflow misses its expected run window
It’s completely free right now. I’m not selling anything — I just want to get it in front of real production users and hear what’s broken, missing, or useless.
If you run n8n in production and want early access, reply here or DM me. I’ll onboard you personally.
Happy to answer any questions about the implementation too.
The cost-per-workflow view is a good idea, especially for AI-heavy client work where the invoice shows up long after the architecture decision was made.
One thing I would add is treating cost anomalies like operational events, not just analytics. For example:
- Cost per successful business outcome, not only cost per execution.
- Alert when cost rises but output volume does not.
- Alert when a model fallback changes cost or latency.
- Link the anomaly to the workflow owner and client.
- Keep a short resolution note: prompt changed, model changed, retry loop fixed, rate limit handled, etc.
The silent-failure side and the cost side are connected. A workflow can look “up” while retrying too often, returning bad output, or spending 4x the expected amount to produce the same result. For agencies, that becomes a maintenance/reporting issue pretty quickly.
This is close to the operating layer I am working on with Maintain Flow: checks, issues, resolutions, and client-ready reporting after workflows are live.