Run many clusters from one screen
Lab, staging and production side by side, with a health dot for each. When one cluster is down, the rest keep working.
Skip to contentPlatform teams Safe self-service On-call Troubleshooting 
New in 2.1: pod shell, YAML editing, sessions Every cluster.
Every cluster.
One calm place.
k10y-agent is an open-source control room for Kubernetes. Monitor, troubleshoot and safely change every cluster from one dashboard, with roles, audit, alerts and an AI assistant built in.
Self-hosted · Free and open source · Your cluster data stays on your servers
Health, problems, CPU and memory for every cluster, updated live.
Works with the Kubernetes you already run
- Kubernetes
- RKE2
- k3s
- Amazon EKS
- Google GKE
- Azure AKS
- Docker Compose
- PostgreSQL
- Telegram
- Google Gemini
- OpenAI-compatible APIs
- amd64 + arm64
2commands to install
27pages in the dashboard
21fine-grained permissions
Everychange audited, with who made it
0kubeconfigs handed out
Get started
Two commands. Five minutes.
Run the hub anywhere Docker runs, then add a small agent to each cluster. No Ingress, no Helm charts, no agents phoning home.
1 · Start the hubDocker Compose: hub + PostgreSQL + automatic backups
curl -LO https://github.com/zorouzumaki1111-cmd/k10y-agent/releases/latest/download/docker-compose.yml
make hub-up # or: create .env, then docker compose up -d2 · Add each clusterOne manifest; the agent creates its own API token
kubectl apply -f https://github.com/zorouzumaki1111-cmd/k10y-agent/releases/latest/download/install.yamlUse cases
Built for the people who keep clusters running
Give developers access without kubeconfigs
A role per cluster: view and logs on production, scale and restart on staging. Developers never hold credentials, and every click is audited.
Get alerted, then fix it from your phone
A crash loop fires a Telegram message, with an optional AI diagnosis. Open k10y, check the live logs, roll back. Done.
Let AI explain a CrashLoopBackOff
The assistant reads events, specs and logs with read-only tools, finds the root cause and proposes a fix. A person reviews it before anything changes.
See
See + understand
- Live charts
- CPU and memory for the cluster, every node and every pod, with request and limit lines.
- Problems first
- Crash loops, failing image pulls, pending pods, NotReady nodes and failed jobs, before you go looking.
- Every resource
- Deployments to PVCs, CronJobs to autoscalers, each with a details panel, related objects, YAML and events.
Control
Control + break nothing
- Dry run everything
- Scale, restart, roll back, drain, apply or edit YAML: see the result before it happens.
- Guardrails
- Protected namespaces, typed confirmation for destructive actions, and conflict checks when editing.
- Least privilege
- Read-only installs are one manifest. Write access is opt-in and can't create RBAC or CRDs.
Respond
Respond + fix fast
- Alerts that wait
- A hold time keeps blips quiet. Reminders, resolved messages, and history in PostgreSQL.
- Live logs and shell
- Follow every pod of a workload at once, or open a shell in a container. Both in the browser.
- AI assistant
- Gemini or any OpenAI-compatible model (optional; cluster data goes to the provider you choose). Read-only tools; you approve every fix.
Everything included
One tool instead of five tabs
Multi-cluster
A cluster switcher with live health dots, and an all-clusters overview.
27 pages
Nodes, namespaces, every workload type, CronJobs, autoscalers, Ingresses, storage, config and events.
Relationship map
Ingress → Service → workload → Pod → ConfigMap, Secret or PVC. Broken links are flagged.
CronJobs
Run now, suspend and resume, with readable schedules like "every 5 minutes".
Per-user settings
Hide pages and set your theme and start page. Your settings follow you to every device.
CLI included
k10y status --all, logs, scale, roll back and apply, against any agent.
Security
Secure by default, not by checklist
k10y sits between people and your clusters, so it was designed to be the safest way in.
Identity
- argon2id passwords, forced first change
- Server-side sessions, sign out any device
- Login rate limiting
Authorization
- Fully custom roles
- Per-cluster or global assignments
- 21 fine-grained permissions
Data
- HTTPS everywhere, no Ingress needed
- Tokens and keys encrypted at rest
- Secret values never leave the cluster
Accountability
- Every action, login and shell audited
- Automatic database backups
- API-only agents, no side doors
Roadmap
Built in the open, shipping fast
Nine releases so far, from the first agent to multi-cluster, the AI assistant, the hub, alerts and pod shells.
Shipped
- Hub with roles and audit
- Alerts to Telegram
- Pod shell and YAML editing
- Live logs
Next
- Two-factor login (TOTP)
- Slack, email and webhook alerts
- Alert silencing
Later
- Single sign-on (OIDC)
- Long-term metrics (Prometheus)
- Helm chart
Bring calm to your clusters today
Free, self-hosted and running in five minutes.