I built a small pre-production review tool for n8n workflows

I work with n8n workflows, and one thing I kept doing manually was the final review before putting a workflow into production or handing it to a client.

A workflow can run successfully, but I still want to check things like:

• What happens if an external write fails?
• Can retries create duplicate rows or messages?
• Is AI output validated before it reaches a business action?
• Is there a real failure path if something goes wrong?

So I built a small free tool called Auturance.

You export an n8n workflow JSON, upload or paste it into Auturance, and it gives you a static review of things that may be worth checking before deployment.

Here is one example.

I scanned a 12-node AI workflow that creates a daily threat intelligence digest.

Auturance gave it a score of 77/100 and surfaced three main areas for review:

• external side effects with no visible failure handling
• a possible duplicate-write risk
• AI output reaching a business write without strong schema validation

What I like about this approach is that the result is not just a score.

You can open each finding and see why it matters, which nodes are affected, and some possible fixes.

For example, in this workflow Auturance noticed that AI output eventually reaches an external write, but the static graph only shows weak validation signals such as Code/IF nodes rather than an explicit structured-output or schema validation step.

That does not automatically mean the workflow is broken.

This is static analysis, so the goal is to highlight things worth reviewing, not to claim that a workflow is definitely safe or unsafe.

Auturance does not run your workflow and cannot prove that credentials, mappings, APIs, AI output, or downstream business results are correct.

There is no account required and no connection to your n8n instance. The original workflow JSON is processed for the scan and is not stored.

I’m mainly looking for feedback from people who build real n8n workflows.

If you try it, the most useful feedback for me is:

• something it flags that is actually fine
• an obvious problem it completely misses
• whether this kind of report is useful before deployment or client handoff

You can try it here:

auturance