~/blog · · Andy Rodriguez

How SetCloud isolates every client's workloads

Organization access, logical database roles, scoped credentials, and the responsibilities that remain in your application.

Architecture description reviewed October 23, 2026.

SetCloud uses organization-scoped access checks and resource credentials on shared infrastructure. These controls protect platform resources; your application must still authenticate its own users and enforce their permissions.

Project databases and credentials

A project database is a separate logical PostgreSQL database with its own login role on a shared private pool. The private connection proxy and pool are shared, while database permissions scope access to each logical database. Platform API operations also check account and organization ownership.

The platform places generated database credentials in an encrypted secret and injects the connection string at runtime. Workload credentials scope access to linked resources. Platform operators require broader administrative access to provision and manage shared infrastructure; this is not a separate cloud account or private network per client.

Secrets and application code

Vault values are encrypted and versioned. Linked runtime secrets stay out of the repository build inputs, but your code can still expose them by logging or returning them. Review application code, permissions, dependencies, and backups as part of your own security work.

A normal secret deletion has a seven-day recovery window; an explicitly forced deletion is permanent. A container-hosting pause retains data and can leave other resources billing. Deleting a project permanently removes its resources and project-managed credentials. Customer-created Vault secrets and their versions remain; the project’s links and access are removed. Delete those secrets separately when no longer needed.

Access and operations

Account Settings offers SMS two-factor authentication. Container terminal sessions use short-lived, single-use tickets and are audited. Project Security offers optional firewall and bot controls for supported deployment types.

These controls do not constitute a security certification or a promise that one client can never affect shared capacity. See Security, Platform Security, and the database guide.