01.Core Principles & Schema-Only Architecture
Blacksmith is engineered as an autonomous reliability proxy for B2B automation workflows. Our core data principle is minimization by design. By default, our healing engine extracts structural JSON schemas (key names, nesting hierarchy, and data types) without transmitting customer values, financial amounts, or private data to external AI models.
- Authentication Stripping: All inbound HTTP headers containing tokens, credentials, or session cookies (
Authorization,Cookie,X-API-Key) are stripped prior to persistence. - PII Redaction: When payload repair requires semantic context, emails, phone numbers, and payment details are sanitized using deterministic client-side token placeholders.
02.Data Retention (72-Hour Ephemeral TTL)
We believe agency client data should not sit in unmonitored databases indefinitely. Quarantined failure records in the Dead-Letter Vault are stored with an automated 72-hour hard Time-To-Live (TTL).
Quarantined webhook bodies are permanently deleted after 72 hours via automated purge routines, providing ample time for agency review without long-term storage risk.
All stored payloads are encrypted using industry-standard AES-GCM-256 before disk write. Keys are isolated per tenant.
03.Third-Party Subprocessors
Blacksmith engages only verified enterprise infrastructure providers:
| Subprocessor | Purpose | Data Handled | Data Location |
|---|---|---|---|
| Supabase Inc. | Tenant DB & Auth | Encrypted events, user accounts | Frankfurt / EU-Central |
| Google Cloud / Gemini | Schema analysis | Structural keys & types (Zero PII) | EU / US Cloud Regions |
| PostHog Inc. | Product analytics | Telemetry events (IPs anonymized) | EU Cloud |
04.Data Protection Agreements (DPA)
Automation agencies processing European or regulated client records may request our standard Data Protection Agreement (DPA) including Standard Contractual Clauses (SCCs).