Instantly see what is failing
The Workloads overview counts every kind and calls out what is wrong: CrashLoopBackOff, ImagePullBackOff, unavailable replicas.
Recent warnings and restarts sit underneath, so the cause is one click away.
Free preview
Stormglass reads your kubeconfig and talks to the API server directly. Lists update as the cluster changes. Logs, shells, files and YAML are one click from any pod. Nothing gets installed in your cluster.
macOS, Linux and Windows. Free during the preview, no card or account needed. Cloud cost is coming soon.

Features
Every built-in resource and every CRD has a list and a detail page. Names are links. Cloud cost is coming soon.
The Workloads overview counts every kind and calls out what is wrong: CrashLoopBackOff, ImagePullBackOff, unavailable replicas.
Recent warnings and restarts sit underneath, so the cause is one click away.
Logs stream as they are written, with levels detected and coloured. Filter by text or level, pick a time range, and download what you see.
One log can combine several pods and containers.
Forget the kubectl incantation. Pick a service or pod, choose the port, and it is on localhost.
Active forwards are listed in one place.
Edit any object in place. Fields the cluster owns are dimmed and left out of the save.
Every change goes to the API server as a dry run first, so admission errors show before anything applies.
Secret values and secret-backed environment variables are blurred everywhere until you ask.
Reveal or copy one value at a time. They are masked before anything reaches the assistant, too.
Scale, restart, cordon, drain, suspend and trigger, from the object's menu.
Each one is checked with a server dry run before it runs.
Stormglass reads your AWS bill and usage, then lists idle and oversized resources by what they cost you each month.
Each finding shows the metric behind it, from GPU utilisation to idle days.
Read-only by default. When a fix is available you see exactly what will run, with a dry run option.
Nothing changes until you press Apply.
Docker
Switch to Docker mode to work with containers, images, volumes and networks on your machine. Containers Kubernetes created are hidden unless you want them.
Assistant
Open the side panel and ask a question. The assistant reads resources, logs, events and metrics with the same access you have, and opens the right page in the app to show its answer.
Turn on the built-in MCP server to give Claude Code, Cursor or another MCP client the same cluster tools. It listens on 127.0.0.1 only and needs a token. Tool permissions are set in the app.
claude mcp add --transport http stormglass \
http://127.0.0.1:17438/mcp \
--header "Authorization: Bearer <token>"
Settings shows the exact command and the Cursor config with your token filled in.
Platforms
Stormglass is written in Rust and draws its UI on the GPU. There is no bundled browser, so it starts quickly and stays responsive with thousands of rows. Themes: Charcoal and Midnight for dark, Paper and Silver for light, with your choice of accent and fonts.
No. Stormglass uses your kubeconfig and talks to the Kubernetes API like kubectl does. Exec auth plugins for EKS, GKE and AKS work as they do in your shell. CPU and memory columns need metrics-server, if you want them.
Any cluster that serves the standard API: managed services, k3s, kind, minikube, OrbStack, Docker Desktop and self-hosted clusters. CRDs are discovered from the cluster, so custom resources show up without setup.
Only what you send to the model you configure for the assistant. If that model runs locally, nothing leaves at all. There is no telemetry and no Stormglass server in between. More on privacy.
It is a desktop app with mouse and keyboard, linked objects, a real terminal and a YAML editor, and it runs locally against your own credentials. It also covers Docker and cloud spend, so one window replaces a few tools.
Yes. Stormglass is in preview and free to use, with no card and no account. Paid plans are planned for after the preview; see pricing.
Download, open, pick a context. No sign-up.