What do you need help with?
Run a complete local Guide: Inspect a configuration · Compose a platform and app · Review a change · Match GPU facts · Recover from a refusal · Save and resume. Each keeps files and results you can review, using cub or an assistant.
You need a configuration, you have one, or you want a whole platform. Start from where you are. Each path gives you exact files and a result you can keep, and the free ones need no account. Read what this site is for the full picture.
1. You need a configuration.
Start from a tested one — a chart at a variant, rendered and checked, with a receipt. Run cub config check <name>, or browse the Catalog's 113 components. Every published version remains available from the public Catalog registry.
2. You have one. Is it right?
Bring your values, YAML, OCI, or work made by AI. The browser check shows the exact objects, the differences that matter, and the checks that did not run.
3. You want a whole platform, not one config.
Compose a stack from tested parts and prove it holds together before any of it runs with cub stack certify. Generate a Kubara platform, deploy your apps on it, and roll back one target without touching its peer.
4. You already run Flux or Argo CD.
Render a reviewed config to a controller-native OCI with one command and no account, then reconcile it the way you do today. Your reconciler applies exactly the published digest, byte for byte. Keep your reconciler.
5. A team needs to share, approve, and promote it.
Upload it into ConfigHub. Release it by digest, promote it from development to production, and gate a release on an approval. Across many clusters, a fleet shows where each one needs attention.
6. You run AI on GPUs.
Inspect a tested AI-platform configuration without a GPU, or compare the GPU nodes you already run. The same review works on an AI platform as on a Helm chart.
This site uses five words in a specific way. A config is one chart.
Apps on a platform defines an app as “a workload you bring.”
Stacks and fleets defines a stack as “a set of parts named in one manifest and checked before any of it runs.”
The same page defines a platform as what a stack becomes once it runs under governance with your apps on it. A fleet is that stack and its apps placed across many clusters.
Upstream moved or vanished? If a chart no longer pulls anonymously, start from a tested successor. If a version now points at different bytes, run the digest-drift check.
You can also see six worked examples, run the short Redis example, or choose a deployment method, or see the four common Helm questions.
Getting started with cub
cub is ConfigHub's command line. The cub config, cub app, cub stack, and cub fleet verbs run today through the workshop plugin, while cub variant (including cub variant promote) and cub release are ConfigHub's own. Add --out oci://… to any free verb to hand its result on as a verified image.
Before you run it, install the cub CLI, then add the plugin verbs with cub plugin install confighub/cub-workshop. The browser check needs nothing installed, and public catalog packages are open to anyone.
See it end to end. Walk it in ten minutes, from one chart to a governed fleet, most of it free and copy-paste.