A Nairotech product · yeke.io

Say it.
See the plan.
Approve it.

Manage Kubernetes clusters from one console. Tell AI what you want to do, review its proposed changes and approve them.

Multi-cluster managementAI assistanceControlled changes
yeke.io · production-k3s · workloads
Deployment list for the selected cluster, grouped by namespace: a kube-system panel, a production panel where api is 3/3 and web-frontend 5/5, and a staging panel below carrying an amber warning stripe
Select a cluster. View workloads by namespace and follow their status live.

In Turkish, yeke is the tiller: the lever used to turn a boat’s rudder. YEKE helps you manage your clusters while you choose the direction.

AI assistance

Work with AI. Stay in control.

Ask about your cluster or request a change. AI works within your permissions, and its proposed operations wait for your approval.

  • Works within your permissions. AI cannot access Kubernetes resources you cannot access.
  • See what is shared. Track data sent to AI. Secret values are never sent to the model.
  • Choose your provider. Use your own provider or a local model with Ollama. Set AI access per cluster.
chat · production-k3s
The YEKE chat surface: the egress transparency card, the identity line, a k8s_list tool-call chip and token and latency telemetry
Ask about your cluster and see which resources AI inspected in the same chat.
Anthropic Claude OpenAI Google Gemini DeepSeek Ollama · fully local Keys are encrypted in the database — not sitting in an environment variable

Change management

Review the plan before you apply.

Whether you create the operation or AI proposes it, the change is checked, presented for approval and recorded.

  1. plan Review the proposed change.
  2. dry-run Validate without applying.
  3. guardrail Check the plan against safety rules.
  4. approval Approve the plan you reviewed.
  5. apply The result is recorded in the apiserver's own words.

If you use hooks, your system is checked when the plan is created and again at approval.

  • Preview the change. If replicas will increase from 3 to 5, review the diff and dry-run result on one card.
  • Check risky operations. Built-in policies evaluate actions such as deletion and permission changes. Risky plans require extra approval.
  • Approve the plan you reviewed. Review again if the plan changes. Supported reverts follow the same approval flow.
plan op-01KZ4M7Q2… · awaiting approval
A YEKE plan approval card: a mutating plan awaiting standard approval, the PATCH request, a before-and-after diff table, a dry-run passed badge and the Apply button
Review the replica change and pre-check results on the approval card.
The same card after it was applied: an Applied badge, the line applied · HTTP 200 · rv 122835 and a Revert button
After approval, follow the result on the same card.
  • Connect your approval process. Use hooks to check conditions such as an open change record or maintenance window.
  • See the decision and reason. Your system can block an operation or require extra approval. The reason appears on the card.
  • Handle connection failures. Plans stop by default if a hook does not respond. You can configure warning-only behavior instead.

Hook setup and examples →

plan op-01KZ7B2R4… · elevated approval
A YEKE approval card raised by a hook's elevate decision: under Verification hooks, the bakim-penceresi hook's elevated badge and its own reason sentence, an elevated approval chip at the top, and at the bottom a confirmation box asking for the resource name to be typed out
The hook decision and any extra approval requirement appear on the same card.
  • Find who did what. Filter operations by user, resource or plan.
  • Track AI activity too. See what it read, what it sent out and which plan was applied.
  • Review the history. Find who changed a replica count in the operation records.
operations · audit trail
The YEKE operations ledger and audit trail: the applied plan row and the ai.read, ai.egress and plan.created events
Follow applied plans and AI activity on one screen.

Monitoring · Scan · KubeWorld

Monitoring is built in. Nothing extra to install.

The YEKE agent collects metrics straight from the kubelet with read-only access. You don't need to install Prometheus, Grafana, node-exporter or metrics-server.

All three are free in the Community edition, with no license key.Alerts, scheduled scans and 90-day history are in Enterprise.

  • Monitoring. CPU, memory, disk and network for nodes, workloads and storage, with 30 days of history.
  • Scan. The “Run scan” button checks the cluster’s health in one pass and reports each finding with its evidence, and with a suggested fix where one applies. It runs on your own AI provider.
  • KubeWorld. A live digital twin of your cluster: namespaces are districts, workloads are buildings, pods are windows. Look at the city and you see where the trouble is.

How monitoring works →

KubeWorld guide →

monitoring · kubeworld
KubeWorld: a cluster city built from demo data. Namespaces are districts (payments, platform, web, data, batch), workloads are buildings; a fire engine outside the building of a crash-looping pod in payments, an ambulance outside the building of an OOMKilled pod in data. On the right, the Selection, Operations and Problems cards.
A fire engine pulls up outside the building of a crash-looping pod, an ambulance for an OOMKilled one; the Problems card in the side panel says why.

Workload status

See the problem in the status view.

See Pod status alongside the error and its duration, including failures such as CrashLoopBackOff.

Rancher · Lens · Yeke

Where does YEKE fit?

YEKE brings multi-cluster management together with AI assistance and human-approved changes.

Capability Rancher Lens Yeke
Visual multi-cluster management ✓the industry standard ✓desktop IDE ✓web, ~390 kB gzipped
Monitoring inside the product —a separately installed Prometheus + Grafana stack —shows metrics if Prometheus is in the cluster ✓node · pod · workload history; neither Prometheus nor metrics-server required
The product names the root cause —you read the graph — ✓“Diagnose” opens with context, “Scan” produces an evidence-backed root-cause report
Footprint installed into the cluster agent + monitoring stackPrometheus, Grafana and exporters, each on its own —a desktop app; installs no components into the cluster one agent60–85 MiB, 4–9 mCPU; measured in three environments (see the note)
Plain-language management — built in — —a separate AI assistant ✓with the context of the screen you are on
Every AI-proposed change goes through approval — — ✓plan → dry-run → 11 built-in CEL policies → typed approval
Your own hook can stop the plan at approval time partlyat the admission layer, not on the approval card — ✓fail-closed by default: if the hook cannot run, the plan stops
What left the cluster, message by message — — ✓“11.6 KiB · 0 paths” plus Secret redaction
Identity verified at the apiserver, credential never centralised its own RBAC layer whatever the kubeconfig says ✓impersonation is asked of the apiserver; in agent mode the credential stays in the cluster
Installing asks for no account, e-mail or key ✓the software is open and free —Lens ID required since 2022 ✓the license is verified offline too
Turkish interface — — ✓TR / EN, one click

This comparison reflects documentation from September 2026; capabilities may differ in newer releases. The numbers in the footprint row are our own measurements of the YEKE agent: 85 MiB on our production cluster, 60 MiB on Harvester, 66 MB on docker k3s (10 Sep 2026, up to four nodes). We put no megabytes in the Rancher and Lens columns: we do not publish a figure we have not measured ourselves. Rancher, Lens and YEKE focus on different needs.

The interface

Light or dark. Your choice.

Work in the theme you prefer. YEKE follows your system’s appearance setting.

production-k3s · workloads · dark theme
The same workload list on the dark theme: the kube-system, production and staging namespace panels, with the status column and the live badge

yeke.io

Try YEKE in your own environment.

Connect your cluster and try your everyday workflows with YEKE. Let’s review what your team needs.