KubeAgent runs in your cluster or locally, watches every pod around the clock, and diagnoses issues through an agentic kubectl loop. Safe fixes apply on their own. Risky ones ping you — Slack, Discord, Teams, Telegram, or PagerDuty.
kubeagent login
kubeagent onboard
kubeagent watch
Watches pods, nodes, and deployments on a continuous tick. Runs in your cluster on its own service account, or locally using your kubectl context. Detects CrashLoopBackOff, OOMKilled, ImagePullBackOff, node pressure, and more.
When an issue is found, KubeAgent fetches logs, events, and your cluster's knowledge base. It traces the root cause through an agentic loop of kubectl lookups.
Configured safe actions can run automatically. Other supported writes pause for approval in the active terminal; unattended runs deny them. Notification channels carry incident alerts.
Intelligent observability on top of your existing cluster. Works with any distribution kubectl can reach.
Read the docs →KubeAgent runs kubectl tools in a loop — logs, describe, events — until it finds the root cause. Powered by Claude AI.
Connect Slack, Discord, Microsoft Teams, Telegram, PagerDuty, or custom Webhooks for incident alerts and recovery updates.
Read-only tools run automatically. Supported writes follow your safe-action policy or require terminal approval; unsupported operations are unavailable.
Onboard once — KubeAgent learns your services, tech stacks, and infra. That context is injected into every diagnosis.
Every incident saved with root cause, fix, and outcome. Feeds back into the knowledge base so diagnoses improve over time.
Install with npx. No agents to deploy, no sidecars, no cluster-wide service accounts. Login works on headless machines via device-code flow.
Connect once in the dashboard. KubeAgent sends incident alerts straight to your team's channel — no polling, no extra tab. Slack can also trigger an AI diagnosis with one click.
One-click AI diagnosis
Server webhooks
Channel connector
Bot notifications
Incident escalation
Send anywhere else
Unlike other monitoring tools, KubeAgent never asks for your Kubeconfig or cluster credentials. It runs locally as a CLI, using your existing terminal context.
╭──────────────────────────╮
│ YOUR NETWORK │
│ kubeconfig · kubectl │
│ │
│ ┌────────────────┐ │
│ │ kubeagent CLI │ │
│ └────────┬───────┘ │
│ │ │
╰────────────┼─────────────╯
│ OUTBOUND ONLY
▼
┌───────────────┐
│ api.kubeagent │
│ alert channels│
└───────────────┘
✓ NO inbound cluster access
✓ NO stored credentials
✓ NO long-lived tokens
Monitoring and alerts are free forever. Every account starts with 200K AI credits for diagnosis and auto-remediation — they never expire. No credit card required. All plans include Slack, Discord, Teams, Telegram & PagerDuty integration and full dashboard access.
kubeconfig to communicate with your cluster. Your credentials never leave your machine or network. AI requests are routed through our servers for billing, but we do not log or store the content. Nothing is retained unless you explicitly opt in for edge-case debugging.kubeagent login, then kubeagent onboard to scan your cluster and build a knowledge base. After that, kubeagent watch starts monitoring. Takes about 5 minutes.