n8n Production Readiness Checklist

Review checklist

Reliability

Error paths existA failure has an intentional destination, recovery action or escalation.Learn more
External writes are retry-safeA timeout or retry cannot silently repeat an irreversible side effect.Learn more
Duplicate execution is consideredWebhook redelivery, manual runs and partial completion have been considered.Learn more
Idempotency is implemented where neededThe workflow can recognize the same business event or safely upsert it.Learn more
API failures are handledTimeouts, non-success responses, rate limits and malformed responses have an expected path.
Timeouts are consideredLong-running or unavailable dependencies cannot leave the workflow in an ambiguous state.
Branches have expected fallback behaviorImportant conditions do not silently skip required work or produce an unreviewed default.

Security

Credentials are referenced safelyCredentials use the intended n8n references and are scoped to the required systems.
Secrets are not hardcodedTokens, passwords and private keys are not embedded in node parameters or exported examples.Learn more
Webhooks have appropriate protectionAuthentication, signatures, gateways or other controls match the exposure of the endpoint.Learn more
Sensitive values are not unnecessarily exposedLogs, notifications and downstream payloads do not reveal more than the next system needs.

AI workflows

Structured AI output is validatedRequired fields, types and allowed values are checked before use.Learn more
Invalid model responses have fallback behaviorMalformed or incomplete output is rejected, retried safely or sent for review.
AI output is not blindly passed into side effectsA model response cannot directly create, send, update or delete without an explicit boundary.
Human review exists where appropriateHigh-impact or ambiguous decisions have an owner and a clear escalation path.

Operations

Logging existsThe team can understand what happened without logging secrets or unnecessary personal data.
Important failures are observableA failure creates a signal that someone can act on within the required time.
Recovery is possibleThe team knows how to replay, reconcile or repair a partial execution.
Critical workflows have an ownerSomeone is responsible for review, incidents, credentials and future changes.

Why each item matters

Reliability protects business state

A workflow can finish with a green execution while a CRM record, payment, notification or database write is incomplete or duplicated.

Security boundaries need evidence

A credential reference or webhook setting is only one part of the control; access scope and upstream protections still need human review.

AI needs a validation boundary

A plausible model response is not the same as a valid business payload. Validate before it becomes a side effect.

Operations make recovery practical

Ownership, observability and replay plans turn an isolated failure into a manageable production event.

Static analysis can help identify structural questions, but a completed checklist still needs workflow-specific testing and an owner who can validate the real downstream behavior.

Use the checklist with an audit

Start with the n8n workflow audit guide to understand the evidence behind common reliability, security and AI validation questions. Then compare the format with the sample report.

Read the production guide

For the full sequence from business outcome to recovery, follow the n8n production workflow guide before using this checklist.

Focused production checks

Error handling and failure paths.
Retries and duplicate side effects.
Hardcoded secret exposure.
Webhook security before side effects.