Security & Trust Center

Every record is keyed and queried per tenant, behind tenant-context middleware. Isolation is enforced at the data layer, not just the application layer.

Data is encrypted at rest with AWS KMS (encrypted persistent state) and in transit with TLS. Tenant secrets and BYO keys are stored via KMS-backed encryption.

Use your own LLM provider key. BYOK spend is recorded at raw provider cost with zero markup — we make your tokens accountable, we do not mark them up.

Every AI dollar is traced to a shipped feature via the Spend-to-Feature Ledger, with receipts, orphaned-spend flagging, and per-team budgets that warn or block before overspend.

Security-relevant actions — run triggers, budget/billing blocks, configuration changes — are written to an immutable, append-only, tenant-scoped audit log, viewable and exportable to CSV inside the app.

Agent execution runs on AWS ECS Fargate with IAM-scoped roles and network controls. Fully per-execution ephemeral, egress-restricted sandboxing is an active hardening item and is not yet claimed as complete.

KMS encryption with per-tenant binding (encryption context) and removal of any non-KMS fallback is being hardened per our internal readiness review before we state it as fully live.

For Delivery Hero EU brands: PII-redacted logging and explicit tenant opt-in before any source leaves the tenant boundary, with a documented data-flow for GDPR and the EU AI Act, are in progress. We do not currently claim certification.

Deploying the platform into your own AWS account is on the roadmap for regulated and data-residency-sensitive customers.

SAML/OIDC single sign-on and SCIM user provisioning (via Amazon Cognito / IAM Identity Center) are planned. Today authentication is GitHub OAuth with tenant-scoped sessions.