WBX Sandbox
More

Security and isolation

Code in a sandbox can be anything – written by an AI, by your users or downloaded from the internet. This is how WBX Sandbox keeps it away from your systems, from other customers and from our infrastructure.

How sandboxes are isolated

  • Every sandbox is a separate Firecracker virtual machine with its own Linux kernel – the same technology that runs serverless functions at large cloud providers.
  • The VM process itself runs locked down (no privileges, its own filesystem view and network namespace), as a second line of defence.
  • Firewall rules block traffic between sandboxes and to our servers, private networks and cloud metadata. Outgoing e-mail (SMTP) is blocked. The same rules apply over IPv4 and IPv6; nothing on the internet can open a connection into a sandbox (its public IPv6 address only answers ping).
  • Every background job gets an execution receipt – what ran, on which template build, with which inputs, and the checksums of its results. WBX can sign it so nobody can edit it later. It is a record of what the platform measured, not proof that the computation could not have been altered. See Signed receipts.
  • Servers and data are located in the EU, operated by KXCOMM s.r.o., an internet service provider.

What you control

  • Network – full internet, an allowlist or no network at all, for the whole team or one sandbox.
  • Secrets – env values of jobs are stored encrypted, redacted in logs and made unreadable when the job ends.
  • Public URLs – protected by a per-sandbox token.
  • Team roles and API keys – see Using the portal.