Browse documentation
Docs/Start here

Platform guide

ComputeCubed documentation

A practical guide to supplying, renting, connecting to, and paying for compute on ComputeCubed.
Product source reviewed: 22 September 2026

Choose a path

How the platform works

ComputeCubed is a two-sided compute marketplace. Providers attach Linux machines to a managed cluster and choose which capacity can be rented. Renters select an advertised allocation and receive either a private workspace or a Kubernetes namespace. ComputeCubed coordinates identity, connectivity, allocation, lifecycle, and billing while the provider retains responsibility for reliable hardware and the renter retains responsibility for their workloads and data.

Provider hardwareValidated listingRental contractPrivate accessMetered settlement

The five objects to know

Cluster

A provider-owned capacity boundary with a declared accelerator family: CPU-only, NVIDIA, or AMD.

Node

A Linux machine joined to one cluster and tracked through connectivity, inventory, benchmark, and readiness states.

Listing

A versioned offer that describes scope, allowed rental modes, resource ranges, prices, and the last date for new contracts.

Rental

A renter’s accepted allocation. Its complete accepted price and end date are recorded at launch and remain fixed for that rental.

Wallet

Rental Balance pays for usage; Earnings Balance receives provider proceeds. The balances are intentionally separate.

A complete lifecycle

  1. 01

    Onboard

    Create an account, accept the current Terms when prompted, add an SSH public key if you will rent workspaces, and complete provider preparation if you will supply hardware.
  2. 02

    Supply

    Create a cluster, join one or more nodes, wait for inventory and benchmarks, then publish a node or cluster listing.
  3. 03

    Discover

    Renters search live node and cluster listings and compare real hardware, uptime, software, network, disk, and benchmark data.
  4. 04

    Allocate

    The renter chooses a supported mode, resources, GPUs, image, and—where required—a pinned node. Capacity is checked again at launch.
  5. 05

    Use and settle

    Access is private over WireGuard. Billing starts when capacity is usable, pauses for detected node outages, and settles when the rental ends.

What is covered

The provider guides cover prerequisites, cluster and node onboarding, validation, listings, pricing, active-rental obligations, delisting, maintenance, earnings, and removal. The renter guides cover marketplace evaluation, workspace and namespace modes, SSH keys, workspace images, WireGuard, SSH tunnels, kubeconfig access, current cost, ending rentals, and reviews. Billing has its own end-to-end guide because money moves through several distinct stages. Selected accounts also have a public API guide for key management and supported automation.

Product behavior changes. The Console remains authoritative for the controls currently available to your account, and the price confirmation shown before launch is authoritative for a new contract.

The website, documentation, and Console open in light theme for new visitors. Use the sun or moon control in the header to switch appearance; the choice is stored in the current browser. Documentation screenshots follow the selected appearance while preserving their original sanitized content.

Was something unclear or out of date?

Include the page title and visible Console version when reporting a documentation issue.