Self-hosted Kubernetes troubleshooting

From “what’s failing?”
to “here’s the evidence.”

Health, logs and capacity in one place. Berth helps small platform teams investigate their cluster, with an optional AI assistant that proposes fixes for you to review.

Get started free See how it works

Runs in your cluster · No account needed to explore · AI is optional

Berth cluster overview showing endpoint health, workload readiness and resource capacity. — Start with the cluster overview. Follow the evidence into the resource that needs attention.

One investigation, fewer tabs

Start with the question.
Keep the cluster in view.

Move from a symptom to the relevant workload, without rebuilding the context at every step.

Find the failing connection

Follow an endpoint through DNS, TLS, routes and backend readiness. See where to investigate next.

Endpoint health

Understand the workload

Bring pod status, events, logs and resource usage into the same investigation. Check the evidence behind a suspected cause.

Explore the dashboard

Plan the next change

Compare requested and used capacity, inspect disruption budgets, and review suggested fixes before applying them.

How AI investigations work

A second set of eyes

Ask a question.
Inspect the answer.

The optional assistant uses read-only tools to investigate the cluster. It shows its tool calls and returns an explanation with suggested commands or manifests.

You decide what to apply. The agent cannot change your cluster.

Explore the AI and data boundaries

Use an Ollama server you operate, or connect Anthropic or Bedrock with Enterprise. Raw AI logs and descriptions are opt-in.

An example Berth assistant investigation showing tool calls and a database availability analysis. — Example investigation · local model · hostnames redacted

Know where it fits

Your cluster.
Your access. Your model.

A workbench for trusted cluster administrators who want a clearer view of what is happening now.

See the full product →

Self-hosted by design: The dashboard runs against your Kubernetes cluster. You control its network access, storage and credentials.
Useful with or without AI: Explore endpoints, events, logs and capacity directly. Connect a model when you want help investigating.
Clear about its limits: Live state complements historical monitoring. Per-user cluster authorization and multi-cluster management are not available.

Start small

Take a look inside your cluster.

Explore the dashboard without a license key. Get a free Community key for changes on clusters up to 10 nodes. Model setup and compute are separate.

Choose your install method →Compare Community and Enterprise

From “what’s failing?”to “here’s the evidence.”

Start with the question.Keep the cluster in view.