TellWang
Dashboard

Isolation tiers

Every Wok is its own private setup, isolated from every other team at the database, network, and authorization layers from the moment it's provisioned. That's the default, and for almost every team it's the right level. The tiers below let you trade density for stronger isolation when a workload or a contract demands it.

What "isolation" actually means here

There are two separate questions, and they're easy to confuse:

Worth being precise about one thing: hardware memory encryption protects everyone on a machine from whoever runs that machine. It does nothing to separate one team from another — that separation comes from the database, kernel, and authorization controls below, which apply on every tier.

The tiers

Standard

Shipped

Your own Postgres, your own services, your own private network — bin-packed efficiently onto shared bare-metal. Strong database-level, network-level, and authorization-level isolation. Teams share the host's operating-system kernel. This is what every Wok runs on today, and it's the default.

Dedicated host

Roadmap

The same stack, but your Wok runs on hardware no other team shares — no shared kernel with a neighbor. For teams that want physical separation without changing how anything works.

Needs: our multi-machine rollout, in progress.

Confidential

Roadmap

Your containers run inside a sandboxed runtime that sits between your app and the host's kernel, shrinking what a hostile workload could ever reach. The strongest separation between teams short of separate machines, while staying dense enough to be affordable.

Needs: a sandboxed runtime enabled on the host. The engineering is in place; we turn it on per machine.

Confidential + hardware attestation

Roadmap

Everything in Confidential, plus hardware that keeps your app's memory encrypted while it runs and can prove to you which code is running. This is the guarantee that we can't read your running memory — for teams whose compliance requires it in writing. Available on confidential-compute hardware, brought online for the deals that need it.

Needs: confidential-compute hardware (AMD SEV-SNP / Intel TDX). Procured per engagement.

A lighter option for small workloads

Micro

Roadmap

For hobby projects and tiny workloads, a lighter tier shares one database across teams using separate schemas, roles, and row-level rules, with a small per-team API in front. It's cheaper because the heavy part is shared — and its isolation is correspondingly lighter, since teams share one database engine. We'll offer it for small, non-sensitive workloads; sensitive data belongs on Standard or higher.

Choosing a tier

Start on Standard — it's the default and it's enough for nearly everyone. Move up when a specific need appears: physical separation (Dedicated), a harder kernel boundary between teams (Confidential), or a written guarantee that we can't read your memory (Confidential + attestation). Pricing for the higher tiers is set per engagement; talk to us about what your workload requires.

i
Standard is live now. Dedicated, Confidential, and hardware attestation are on the roadmap — this page will mark each as shipped the day it goes live, and never before.