Isolation
- k3s
- Full
- k0s
- Full
- CAPI
- Full
- vkobe
- Namespace
Pre-warmed virtual clusters for your pipelines. Claim one in under 5 seconds, use it, release it.
Define once, claim instantly, release cleanly. Kobe handles provisioning, recycling, and pool maintenance in the background.
Declare the runtime, size, and warm count. Kobe provisions and maintains a ready pool of clusters automatically.
Your CI job uses its existing identity: a GitHub Actions token, Clerk, or Auth0. No API keys, no secrets to rotate.
One request, via the HTTP API or the kobe CLI, returns a kubeconfig once the lease binds. Pre-warmed pools bind in seconds. Run your workloads immediately.
When the job finishes or the TTL expires, Kobe recycles the cluster and refills the pool. No manual cleanup.
Declare the runtime, size, and warm count. Kobe provisions and maintains a ready pool of clusters automatically.
Your CI job uses its existing identity: a GitHub Actions token, Clerk, or Auth0. No API keys, no secrets to rotate.
One request, via the HTTP API or the kobe CLI, returns a kubeconfig once the lease binds. Pre-warmed pools bind in seconds. Run your workloads immediately.
When the job finishes or the TTL expires, Kobe recycles the cluster and refills the pool. No manual cleanup.
Policy-as-code authorization, golden snapshots, autoscaling pools, and built-in observability. Everything you need for production CI/CD infrastructure.
Pre-warmed pools keep clusters running, so a claim binds in seconds. The vkobe backend provisions in about 5 seconds.
Authenticate with OIDC, SSH keys, bearer tokens, or ServiceAccount tokens. Works with GitHub Actions, Clerk, and Auth0, with no long-lived secrets to rotate.
AccessPolicy CRDs give fine-grained control: per-identity max TTL, concurrent-lease limits, and pool allowlists.
Velero snapshots a fully-configured k3s or k0s cluster and restores pool members in seconds instead of cold-provisioning from scratch.
Pools scale between minReady and maxClusters with scale-up thresholds and queue timeouts, so warm clusters are ready before CI asks.
Prometheus metrics at /metrics, plus diagnostic bundles captured on lease release or expiry and uploaded to S3-compatible storage.
Pluggable backends with no vendor lock-in. From lightweight virtual clusters to full CAPI-managed infrastructure.
kobe's own single-pod virtual cluster runtime. The lowest overhead per cluster, with no third-party dependencies.
Lightweight Kubernetes run as StatefulSets. The default backend: fast to provision, low overhead, well-tested.
The k0s distribution, same pattern as k3s. For k0s-specific features or standardizing on k0s in production.
Delegates provisioning to any Cluster API infrastructure provider for full lifecycle control.
The kobe CLI is a single signed binary. Install it, point it at your operator, and lease a fully isolated cluster in seconds.
mise use -g github:kunobi-ninja/kobe@latestbrew install kunobi-ninja/kunobi/kobecargo binstall kobectlkobe config set endpoint https://kobe.example.comkobe loginkobe lease ci-smallKUBECONFIG=~/.kube/kobe-<id> kubectl get nodesKunobi Kobe is open source and Apache-2.0 licensed. Run the operator in your own cluster and never block a pipeline again.