Let’s talk
GUIDE

Who can fix a broken n8n workflow, and what does a repair cover?

Anyone who can read the failed executions in n8n can find why a workflow broke; what you are paying for is the fix holding afterwards. A good repair finds the cause in the execution that failed, fixes it, tests it with the input that broke it, and adds an alert so the next failure is seen the same day.

Most broken workflows fail for a handful of reasons: an expired or revoked credential (401 or 403), a trigger that fires twice for the same thing, an outside service that changed its response, a step that receives an empty value and carries on, or a run that stops without anyone hearing about it.

Updated Oct 8, 2026

How it works

  1. Read the failed execution

    n8n keeps past executions: the one that failed shows the node, the input it received and the error. That is the starting point, not a guess from the canvas.

  2. Reproduce it

    Re-run the failing step with the same input, so the fix is tested against the real problem.

  3. Fix the cause, not the symptom

    Renew a credential the right way, make a step skip what it already handled, check a value before it is used, or handle the outside service’s new answer.

  4. Add an error workflow

    An error workflow with the Error Trigger runs whenever the workflow fails and can email or message someone with the execution link, so a failure is seen the same day.

  5. Hand over a short note

    What broke, why, what changed and how to undo it, so the next person does not start from zero.

What it costs to run

A repair is paid time, not a list price, so there is no third-party price to show here. Running n8n itself costs what your plan or server costs; a repair does not change that.

What we learned building it

Our own n8n workflows (lead capture, reply triage with approval, old-lead reactivation), live on n8n Cloud since 25 September 2026. Not client systems.

A waiting execution holds old data
A run paused on a Wait node kept the values it had when it paused. We moved the state that changes into an n8n data table, read fresh when the run resumes.
A form can send empty answers
Answers from a Wait form are keyed by the field label, not its name, so a step reading the name got nothing and carried on. We now check the value before using it.
One alert per lead needs a fingerprint
Without a check, the same lead could alert twice. A fingerprint of the email or phone, kept for 30 days, stops the repeat.

When it is not worth it

  • If the workflow is small and rarely runs, rebuilding it cleanly can cost less than diagnosing it.
  • If the outside service the workflow depends on is being retired, fix the dependency first; repairing the workflow around it will not last.

Questions

Do I need to give full access to my n8n?

Access to the workflow and its executions is enough to diagnose. Credentials stay yours; a repair can ask you to reconnect one instead of receiving the password.

Can it be fixed under my agency’s name?

Yes. The note and any message to your client can go out as your agency; the workflow and the client stay yours.

How do I know it will not break again?

You cannot be sure, but you can make sure you hear about it: an error workflow sends the failure the same day, with a link to the execution.

Sources

Have a workflow that stopped working, for your business or a client?

See the service: For agenciesLet’s talk

All guides