Organisation isolation
Every company, assessment and recommendation is tied to an organisation, and access is checked per request instead of assumed from the UI.
SECURITY MODEL
Security here comes from boundaries, not from inflated promises.
AI-CRF Monitor deliberately separates data access, privileged operations and scoring logic. The frontend stays static, while security boundaries live in database policies and a small number of focused backend functions.
Row Level Security
Supabase RLS is the access-control boundary. Client-side route guards improve usability but are not the authority for entitlements or data access.
Minimal privileged backend
Only Stripe, invitation and selected administrative operations use Cloudflare Pages Functions with server-held secrets.
No generative-AI scoring
Financial inputs are not sent to an LLM to calculate scores, fill missing values, rank priorities or improvise recommendations.
Data minimisation
The product requests only the data required by the documented score model and the resulting audit trail.
Auditability
Each assessment stores the formula version, input snapshot, derived metrics and triggered recommendation set.
Transport and hosting
The intended deployment path uses HTTPS through Cloudflare and Supabase, without claiming certifications or residency guarantees that have not been verified.
Honest limitations
No digital system can promise absolute security. Risk is reduced through conservative design, testing and limited privilege, not marketing language.
Start on Free
Start with a defensible baseline.
Create your first company, enter the key financial values and receive a reproducible assessment.