Hi n8n community,
I’m opening a controlled beta for PII Redactor, a community node designed to reduce the exposure of structured personal data in LLM-based workflows.
The intended workflow is:
Webhook
→ PII Redactor: Redact
→ OpenAI / Anthropic / Gemini
→ PII Redactor: Restore
→ Response
Redact replaces supported values with readable placeholders:
synthetic.user@example.com
→ {{EMAIL_1}}
The model receives only the masked text. Restore uses the requestId from Redact to recover the original values when the placeholders remain unchanged.
Supported in the current beta:
- Email addresses
- International and local phone numbers
- Luhn-valid payment cards
- IPv4 and IPv6 addresses
- CIDR ranges
- IBANs
- Contextual national identifiers for 18 countries
Current implementation:
- n8n community node with Redact and Restore operations
- Managed HTTPS API
- Per-client API keys
- Temporary Redis mappings
- Configurable TTL
- Single-use Restore
- Atomic idempotency
- Client isolation
- Key rotation without downtime
- Stable error responses
Current validation:
- Tested with Redis 7.4
- Redact and Restore tested through HTTPS
- Packaged node installed in a clean n8n instance
- No known vulnerabilities in the current audit
Limitations:
- Best-effort detection, not complete anonymization
- No names, postal addresses, or organizations yet
- Local phone detection depends on phoneCountry
- National identifiers require context
- Modified placeholders cannot be restored
- n8n execution history must still be configured securely
I’m looking for 3–5 self-hosted n8n users for a 1–2 week beta using synthetic or controlled data.
I’m particularly interested in feedback about:
- Node installation
- Credential setup
- Preserving requestId around AI nodes
- Multi-item workflows
- Placeholder preservation across different LLMs
- False positives and false negatives
Project page: