n8n Workflow: Redact & Restore PII with Guardrails

Mask PII before an LLM call—then automatically restore it with reversible tokens

This n8n workflow demonstrates a safe, end-to-end pattern for redacting sensitive personally identifiable information (PII) using Enterprise PII Guardrails, sending the sanitized text through a simulated LLM step, and then restoring the original values in the final output using a token-to-original mapping.

What this workflow does

• Manual test run: Start the workflow manually to process a sample customer support prompt.
• Sample PII creation: It builds a customer support message containing common sensitive data such as name, email, phone number, medical record number, and a credit card number.
• PII masking with guardrails: The prompt is sent to Enterprise PII Guardrails to detect sensitive entities and replace them with surrogate tokens. The workflow also receives a token-to-original mapping.
• LLM-safe processing (simulated): A simulated LLM response is generated that references the surrogate tokens rather than the original PII.
• Unmask and restore: The simulated response and the saved mapping are sent back to Enterprise PII Guardrails to unmask tokens and restore the original PII in the final text.

Use cases

• Test and validate an LLM prompt-sanitization workflow for customer support tickets.
• Reduce PII exposure when building LLM-powered assistance while still returning accurate, real details to users.
• Provide a concrete template for implementing reversible redaction with Guardrails and an n8n-friendly token mapping flow.

Technical details

• Workflow nodes: manual trigger, set, sticky note, and n8n-nodes-piiguardrails steps around mask/unmask.
• Optional LLM integration: The example uses a simulated LLM step, but you can replace it with your real LLM node (the key requirement is keeping mask/unmask steps before and after).
• Enterprise PII Guardrails setup: install the community node (n8n-nodes-piiguardrails), run Guardrails locally (e.g., pip install piiguardrails && piiguardrails start), and point n8n to your Guardrails endpoint (default http://localhost:8000) via an API credential.