Security & Governance

Trust to delegate — even what is sensitive.

Finance, health, legal: the processes most worth delegating are the scariest ones. That is why security here is not a compliance afterthought — it is five rules in the path of every action your app executes. This page shows what backs each one, and is honest about what is not ready yet.

The trust rules

What makes the operation trustworthy — in five rules

Not audit jargon. These are the conditions for a non-technical person to delegate a real operation — and every product decision in the platform answers to them.

01

Framing

solves the right problem

02

Coherence

makes sense, end to end

03

Fix + undo

mistakes don't hurt

04

Human takes over

a person decides the sensitive cases

05

Your data, yours only

truly isolated

This is what opens the most sensitive domains — finance, health, legal.

Below, what backs each rule: the way apps are built (rules 1 and 2), governance in the path of every action (3 and 4), and the isolation-and-proof engineering (5).

Rules 1 and 2 · Framing and coherence

You watch the app being born — and check before you trust

Your app doesn't arrive in a black box. It builds itself from your problem, in front of you — and every conclusion arrives with proof. Wrong framing shows up on the first screen, not months into a project.

Solves the right problem

Building starts from the description of your problem, and you follow every step on screen. What doesn't make sense for your business, you adjust by talking — no tickets, no waiting for a sprint.

Makes sense, end to end

Not a slice that pushes the rest off to later: the solution is born whole and coherent, and the conclusion of each stage comes with the evidence for you to check.

Create an app by talking(Live interactive demo — no sign-up needed)

Rules 3 and 4 · Mistakes don't hurt, a human takes over

The AI truly executes — inside your rules

The difference between an AI that helps and one that scares is who decides. Here, governance is not a separate dashboard: it sits in the path of every action your app executes — so that mistakes don't hurt and, in sensitive cases, a person decides.

Human in the loop (HITL)

Tiered approvals: nothing critical executes without the OK of whoever is in charge. The AI proposes; the decision that matters stays with you.

Policies and roles

RBAC and policies decide who can do what — enforced by the platform at every step, not trusted to each app’s code.

Quotas and usage limits

Usage metered and gated per tenant, with atomic reservation. No silent overruns, no consumption surprises.

Consent and privacy

GDPR/LGPD consent lifecycle with a trail — legal basis recorded, not presumed.

Partial — advanced governance maturing
Rule 5 · Your data, yours only

What protects your data — and how to prove it

Four foundational protections, active from day one — each with the engineering trail that backs it.

Isolation

Your data stays yours only

Each company has a dedicated space. No customer ever sees another’s data — not by accident, not on purpose.

Dedicated schema per customer + RLS as backstop. The tenantId derives from the request context, never from the payload.

trail: securityEngine · dataEngine
BYOK

The key is yours — and so is the lock

Customers who require it can bring their own encryption key. Revoke it and the data becomes unreadable immediately — with or without us in the loop.

BYOK with KMS (AWS/GCP/Azure) and crypto-shredding: revocation cuts access to the data, independently of the platform.

trail: securityEngine · spec-byok
Privacy

What is sensitive never reaches the model in raw form

Personal data is masked before any AI processes it — and there are guards against attempts to manipulate the model.

PII masking (4 strategies) before the LLM + content safety against prompt-injection and jailbreak.

trail: securityEngine · spec-content-safety
Audit

Every action recorded, tamper-evident

Who did what, when — in a trail that cannot be edited without leaving a mark. Verifiable by you, not just on our word.

Dual-token authentication + RBAC; SHA-256 append-only trail — tampering is cryptographically detectable and verifiable per tenant.

trail: securityEngine · auditTrailEngine · spec-hash-chain
Implemented — verifiable in the architecture today
Compliance, honestly

What is ready — and what is not yet

The platform has automated scanners for SOC2 controls (CC6/CC7/CC8) and GDPR (Art. 17/20/30/32) — verification by code, not a manual checklist; today they run on demand, and scheduled continuous execution is on the roadmap. We do not yet have a formal third-party SOC2 certification or a completed external pen-test — they are the next step, not a promise fulfilled. What we claim above is what can be verified in the architecture today.

FrontWhat we already deliverStatusNext step
Data protectionMulti-tenant isolation, BYOK, PII masking, hash-chain audit ImplementedExternal pen-test
GovernanceHITL, policies/RBAC, quotas/entitlements, consent PartialBroad fail-closed + more autonomy patterns
ComplianceAutomated SOC2/GDPR scanners (on demand) PartialContinuous execution + SOC2 Type II certification (external audit)
Deep evaluation

Going to pop the hood on security?

For those who need to evaluate in depth, we provide an isolated evaluation environment and an architecture session with engineering — where every guarantee above is demonstrated, not asserted.