M
Security
Security is a product feature.
Per-brand isolation, encrypted tokens, TLS everywhere and a real incident runbook.
Data isolation
Every brand gets its own RAG index, storage prefix and row-level scope. Cross-brand reads are impossible by policy, not convention.
Encryption
AES-GCM for third-party OAuth tokens with per-project keys. TLS 1.2+ for transport. Postgres at rest is encrypted by the managed provider.
Access control
Roles live in a dedicated table with a SECURITY DEFINER check. No client-trusted admin flag anywhere.
Auditability
Every agent run, mutation and OAuth event lands in an append-only log with actor, IP and outcome.
Incident response
Sev-1 target: 15-min ack, 4-hr customer comms, post-mortem within 5 business days.
Data rights
Export as PDF, CSV and JSON in one click. Full delete pipeline in Privacy tab and via support.