Security
Isolation by default.
Every project runs in its own network, with its own database and encrypted credentials. The boundaries hold, so a failure in one tenant can't reach the next.
Security questionnaires and architecture reviews get a reply from cloud-claude.com.
microVM
Hardened Firecracker sandbox per service
AES-256-GCM
Secrets encrypted at rest
OAuth 2.1
PKCE S256, mandatory for agents
Per tenant
Own network, own database, ownership-checked
Isolation
Every service runs inside its own boundary.
A hardened microVM per service. A private network and an encrypted database per project. Ownership checked on every request.
microVM sandboxes
A hardened Firecracker microVM per service.Isolated networks and databases
Each project gets its own Docker network and its own Postgres database with its own credentials.Ownership checks
The dashboard, chat, and MCP all verify ownership before they act.
Identity & data
Humans and agents authenticate the same way.
A person in the dashboard and an agent over MCP run under one permission model and one audit trail. Secrets are encrypted before they are stored.
OAuth 2.1 for agents
PKCE S256 is mandatory on every authorization.Encrypted secrets
AES-256-GCM at rest, injected through a tmpfs file and never in a process argument.Honest posture
We do not call this zero-knowledge, because it is not.
No certification yet — SOC 2 readiness is on the roadmap.
Report a vulnerabilityBring us your questionnaire.
Security questionnaires, architecture reviews, and evaluations get a reply from cloud-claude.com. Or join the waitlist and evaluate it yourself.