Mental model

Komputo is best understood as git push for compute jobs, not an AWS console. You hand it work; it decides where the work runs.

   you describe          Komputo picks the            you get back
   the work       ──▶    best EU place &      ──▶     outputs + a receipt
   (a job)               runs it for you              (cost · CO2 · EU residency)

The five nouns below are everything you need to hold in your head.

Jobs

A job is a unit of work: a container image, a command, the resources it needs, and a policy. You submit it; Komputo decides where to run it. A job moves through a canonical set of states from QUEUED to COMPLETED (see Job states).

Providers

European compute underneath the scheduler, in three forms — your connected cloud accounts (BYOC), your self-hosted runners, and the free shared pool:

            ┌─────────────────────────┐
   a job ─▶ │  Komputo picks one of:  │
            ├─────────────────────────┤
            │  ☁  your connected cloud │  your account, your rates
            │  🖥  your own runner      │  a machine you attached
            │  ◇  the free pool        │  metered EU hosts, no setup
            └─────────────────────────┘

You normally ignore which one runs a job — Komputo picks the best place right now. Power users can pin a provider or set placement: prefer-self-hosted.

Organizations

The access and attribution boundary. Jobs belong to a project inside an organization; sovereignty policy, budgets (scheduling spend-ceilings, not money), and attribution live at this level. Komputo takes no payments — providers bill you directly.

Receipts

Proof of what a completed job cost, how much energy it used, how much CO2 it emitted, and that it ran inside the EU. A receipt is informational — evidence for showback and CSRD reporting, never a bill. It’s the reason teams trust the numbers.

Artifacts

The outputs a job produces, persisted to EU storage. Browse and download them after the job finishes — nothing is lost when the node is torn down.