// Kubernetes incidents, diagnosed
Your AI SRE for
Kubernetes.
Kubently Cloud connects to your clusters with a read-only, outbound-only agent and diagnoses crash loops, failed deploys, and pipeline errors — automatically from Alertmanager, or on demand in plain English.
No card required · read-only access · one helm install
checkout-db, which no longer has a DATABASE_URL key — renamed in the last release.Nothing to open up.
The executor dials out to the control plane, so your cluster never accepts an inbound connection — and never hands over a kubeconfig.
Built for real incidents
One helm install
A lightweight executor dials out to Kubently Cloud. No inbound firewall rules, no kubeconfig upload, read-only by default.
Agentic diagnosis
An AI agent runs real kubectl investigations — multi-step, across clusters — and explains the root cause, not just the symptom.
Alert to answer
Point Alertmanager at your private webhook and get root-cause analyses delivered to Slack before you've opened your laptop.
Read-only, tenant-scoped
Every command is validated and RBAC-scoped in your cluster, and the control plane scopes each API key to its own tenant's clusters.
Speaks A2A and MCP
Chat from the dashboard, the Kubently CLI, or any A2A- or MCP-compatible agent — same read-only cluster access.
Built on the open core
The in-cluster agent is the open-source Kubently executor, pinned by version. Nothing proprietary runs in your cluster.
Three steps, about a minute
Sign up
Google, GitHub, or an email link. No card required.
Run the generated command
One helm install in any cluster wires up the executor with your token.
Ask, or wire up alerts
Chat with your cluster from the dashboard or CLI, and point Alertmanager at your webhook for automatic triage.
Stop reading logs at 3am
One helm install and your next crash loop arrives already diagnosed.