Security

How Codegenix isolates tenants, sandboxes code execution, and protects your secrets.

Tenant isolation

Every workspace's data lives behind Postgres Row-Level Security. Each workspace's auto-provisioned backend is a separate Supabase project, so one tenant's tables are never reachable from another's.

Sandboxed execution

Generated apps run in StackBlitz WebContainers (an in-browser runtime) or E2B Firecracker microVMs — never directly on shared infrastructure. The agent worker bounds every run with wall-clock, concurrency, and cost caps.

Secret handling

Bring-your-own-key provider credentials are encrypted at rest with AES-256-GCM under a versioned key id. Server-only secrets never carry a NEXT_PUBLIC_ prefix; the structured logger and the error tracker scrub secrets and PII through one shared deny-list before anything leaves the process.

Request hardening

Mutating API routes enforce same-origin (CSRF) checks, authentication, workspace role, and rate limits, and write an audit row. Responses carry HTTP Strict-Transport-Security, a Content-Security-Policy baseline, and a restrictive Permissions-Policy. The agent's web-fetch tool is SSRF-guarded and pins its connection to a pre-validated IP.

Reporting a vulnerability

Email security@codegenix.dev with a description and reproduction steps. We aim to acknowledge within three business days. Please do not open a public issue for security reports.

Compliance

Codegenix is pre-general-availability. Formal certifications such as SOC 2 are on the roadmap and are not yet attained — this page describes current engineering practice, not a certification claim.