OrkStageOrkStage

Introduction

OrkStage is an open source multicloud control plane. Define an environment once, place each resource on a cloud, review the plan and deploy with OpenTofu.

Alpha

OrkStage is in alpha. The whole platform is being tested and validated, and features and provider integrations are released gradually. APIs and data formats may still change between versions; keep that in mind before you run anything you depend on. See roadmap for what is planned.

OrkStage provisions and operates isolated environments on one or several clouds. Each environment is called an instance (or a stage). It is a logical unit that groups services, a Postgres database, a cache, a queue, object storage, ingress and DNS.

OrkStage places each resource on its own. One instance can run its API on AWS, its database on Azure and its storage on OCI.

Every instance follows the same flow:

  1. A template describes what the environment needs, without naming a provider. You write it once and deploy it many times.
  2. An instance binds each placement slot of the template to a cloud target and a region.
  3. The plan resolves every resource, evaluates policies, estimates the monthly cost and picks a connectivity strategy for each link between clouds. No cloud is touched at this step.
  4. A deployment applies the plan in the background with OpenTofu, following the dependency graph.
  5. Health checks roll up to the instance. They open incidents, update a private status page and send notifications. Every change is recorded in an audit log that only accepts inserts.

What works today

OrkStage provisions real AWS resources with OpenTofu. We test this locally against MiniStack, and it works on a real account through a role that OrkStage assumes in that account. Azure, Google Cloud, OCI, DigitalOcean, Vultr, CoreWeave, Magalu Cloud, Vercel, Render and Supabase have adapters in Preview: covered by tests and tofu validate, not yet proven on live accounts (see providers). The simulation provider imitates any of them without an account and labels its resources as simulated. The ork CLI is in Preview: it works against the API, but signed binaries are not released yet.

Where to go next

Who it is for

  • Platform and DevOps teams that run the same environment shape for many customers or teams. They get one reviewed path from template to cloud, with policy checks on every plan.
  • SREs who need health, incidents, status and an audit trail tied to the infrastructure that was actually deployed.
  • Engineering managers who approve production changes. They see the plan, the risk and an approximate cost before anything is created.

Open source

The core is licensed under the GNU AGPL-3.0 and is complete on its own. You can run it on your own infrastructure and use it for production workloads on AWS. Enterprise features are under a commercial license: the official images include them, and an offline license key turns them on. See Community vs Enterprise.

On this page