KubeWatcher is an AI on call assistant for Kubernetes. It watches the systems your apps run on, tells you in plain English what broke and why, and suggests the fix. Nothing changes until someone on your team says yes.
Free plan. No card needed.
A service keeps restarting. KubeWatcher explains that it runs out of memory, suggests giving it twice the memory, a person approves, and the service is healthy again.
Less noise, an answer you can check, and a fix that waits for you.
| Today | With KubeWatcher |
|---|---|
| Someone is woken up and starts digging through logs | The cause is already written down, with the evidence |
| Twenty alerts for one problem | One incident per broken service |
| The fix is typed in by hand at 3 a.m. | A checked fix waits for one click, and can be undone |
Checkout service keeps restarting
8 alerts grouped, with their full history.
Three steps, and you decide at the one that matters.
Watch · explain · approve
Step 01
Install one small agent in a few minutes. It only calls out to us, so no ports are opened.
Step 02
When something fails, KubeWatcher groups the alarms into one incident, finds the cause and shows its evidence.
Step 03
You see the exact change. Approve it and it is applied, checked and recorded.
It can look before it touches, and it never touches without you.
It can look but not touch. Allowing changes is your switch, in your own cluster, and we cannot flip it.
How?Edit the fix and the approval resets, so what you approved is exactly what runs.
How?Each change is saved before it happens, so going back is one more approval.
How?What it touches, how it talks to us, and what happens when it changes something. Send this section to the people who will install it.
The same incident, as you will see it
checkout-7d9f4 · CrashLoopBackOff
− memory: 128Mi
+ memory: 256Mi
helm install kubewatcher-agent oci://registry-1.docker.io/aaraz/cluster-agent --version 0.1.1 -n kubewatcher-system --create-namespace -f https://kubewatcher.com/v1/enroll/<token>/values.yamlThe dashboard gives you this command with your own one time link in place of <token>.
Start free. Upgrade once when you need more, and buy AI credits only when you use them.
Try it on a small setup.
$0
For teams running real workloads.
Custom to how you work.
Contact us
Choose when the AI runs: automatically, only when asked, or never. If credits run out, alerts and approvals keep working.
Connect a cluster in a few minutes. It watches, explains and suggests the fix. You stay in control.