KubeView — Kubernetes, in full view

v0.1.0 · Milestone 1 shippedSee what's next

A local-first Kubernetes desktop IDE with live telemetry, streaming logs, interactive exec, and gated GitOps drift. Direct from your kubeconfig, with no agents, no accounts, and no telemetry.

Get KubeView

See the architecture

macOS · Windows · LinuxTauri 2 · <80 MB RAMNo account required

Cluster overview

prod-us-east-1 · connected3 warnings

CPU

62%

48.6 / 78 cores

Memory

71%

181 / 256 GiB

Nodes

6 / 6

all ready

Node matrix

4 of 6 nodes · 1 pressure

ip-10-0-1-14us-east-1a64%22 pods

ip-10-0-2-37us-east-1b51%19 pods

ip-10-0-3-88us-east-1c78%24 pods

ip-10-0-1-92us-east-1a43%18 pods

JSON-RPC 2.0 · stdioengine ready

Engineered on a serious stack

local-first · agentless

TAURI 2 · RUST · GO · CLIENT-GO · REACT 19 · NEXT.JS · TAILWIND CSS 4 · SHADCN/UI · MOTION

LIVE TELEMETRY · WATCH STREAMS · PTY EXEC · GITOPS DRIFT · ZERO AGENTS · COMMAND PALETTE

A real moment

One cluster. Three minutes.

Follow a pressure event from the overview to a gated fix. The workspace follows the story instead of resetting at every step.

02:14overview

Pressure appears.

A node crosses 91% memory. The overview shows the pressure, the workloads, and the namespace it is affecting.

02:15pods

Read the pod.

Open the workload behind the warning. Restarts, phase changes, and the surrounding pods are already in the same view.

02:16gitops

Review the fix.

Compare desired and live state, check the dry run, and apply behind a gate. The context never leaves the workspace.

See the full workbench

KubeView — prod-us-east-1

Cluster overview

connected

prod-us-east-1 · 6 nodes · 128 pods · all namespaces

Search⌘K

all namespaces

watch stream live

CPUlive

62%48.6 / 78 cores

Memorylive

71%181 / 256 GiB

Nodeslive

6 / 6all ready

Workloadslive

12824 deployments

Node matrix6 ready · 1 pressure

ip-10-0-1-14

64% cpu22 pods

ip-10-0-2-37

51% cpu19 pods

ip-10-0-3-88

78% cpu24 pods

ip-10-0-1-92

43% cpu18 pods

ip-10-0-2-11

69% cpu21 pods

ip-10-0-3-45

38% cpu24 pods

Warning events3 new

BackOff

pod/api-7d9f8c5b6

FailedScheduling

pod/worker-2b7d4

Unhealthy

pod/web-5c8d7

event stream connected

Scroll the story · or click a tab

The product

Every surface answers the 2 a.m. question: what changed, where, and why. No dashboards to configure. No agents to install.

Overview

Live telemetry, not last-minute polling

CPU, memory, node conditions, and workload counts as a bento matrix, scoped to any namespace in one click.

Explore feature

CPU allocation62.1%

48.6 cores used78 allocatable

Memory

71%

181 / 256 GiB

Nodes

6 / 6

all ready

Pods

128

+4 today

Deploys

healthy

Real-time

Watch streams keep the grid honest

Informer-backed watches push pod changes the moment they land. No refresh button.

Logs and shell, where the pod is

Tail container logs or open an interactive PTY without leaving the inspector.

Explore feature

api-7d9f8c5b6 · /bin/sh

$ kubectl logs -f api-7d9f8c5b6-x2k4p

2026-10-28T20:14:02Z INF server listening :8080

2026-10-28T20:14:03Z INF connected to postgres

2026-10-28T20:14:04Z INF readiness probe ok

Inspect

The whole object, not a summary

Conditions, events, env vars, probes, volumes, and live YAML in one drawer.

Explore feature

apiVersion: v1
kind: Pod
metadata:
  name: api-7d9f8c5b6-x2k4p
  namespace: default
status:
  phase: Running
  podIP: 10.0.1.44

YAML Events Conditions

Connect

Every kubeconfig, one hub

Auto-detect ~/.kube, drag in folders, paste YAML, and run four-stage diagnostics.

Explore feature

~/.kube/configdetected

~/clusters/prod.yamldetected

eks-prod · us-east-1connected

NetworkAuthTLSRBAC

GitOps

Drift, diffed. Apply, gated.

Compare local manifests with live resources, review the diff, then apply only after a dry-run and explicit confirmation.

Explore feature

deployment.yamlchanged

service.yamlmissing

configmap.yamlin sync

Dry-run passed · 2 warnings

# deployment.yaml

- replicas: 3

+ replicas: 5

image: ghcr.io/acme/api:2.4.1

strategy: RollingUpdate

gated by confirmationReview

Keyboard

Command palette first

Jump to any resource, switch contexts, or trigger actions with ⌘K.

Switch context: staging

Open GitOps drift

Local-first by construction

Your credentials never leave the engine.

KubeView connects straight from your workstation to the Kubernetes API server. The Go engine resolves certificates, tokens, and exec auth plugins in memory. The React UI never touches a credential, and nothing is installed in your cluster.

Credentials stay in Go memory

Tokens and certificates are resolved inside the engine — never passed to the React UI.

stdio, not a TCP port

Host and engine speak over private pipes using versioned JSON-RPC 2.0. Nothing listens on localhost.

Agentless by default

No operators, CRDs, or sidecar workloads are injected into your cluster.

In-cluster agents

Cloud accounts

Open TCP ports

Operations local

Architecture

Three layers.
One clean trust boundary.

The UI renders. The Rust host gates. The Go engine talks to the cluster. Credentials never cross into the React layer, and the engine only answers explicitly named operations.

No open TCP port on your workstation
No in-cluster agents or operators
No remote backend or telemetry

Nothing from the frontend can execute an arbitrary command. Every operation is named, validated, and forwarded over private stdio pipes.

Interface

Next.js · React · Tailwind

Displays state and requests operations. Ships as a static bundle inside Tauri and never owns a cluster credential.

Static exportNo credentialsZero Node runtime

Native host

Tauri 2 · Rust

Supervises the Go sidecar, enforces host capabilities, and exposes a narrow, allowlisted IPC gateway.

Capability checksSidecar supervisorOS keychain

Kubernetes engine

Go 1.23 · client-go

Resolves kubeconfig auth, runs informers and watches, and executes every cluster operation over client-go.

JSON-RPC 2.0Informer watchesPTY & logs

Cluster

Kubernetes API server

The authoritative gatekeeper for state and RBAC. KubeView respects the permissions you already have.

RBAC enforcedTLSWatch · SPDY

Roadmap

Shipping in the open.

Milestone 1 is complete and used daily. Here is what lands next, in the order it will land.

Phases complete6 / 16

Completev0.1

Core IDE

The daily-driver foundation is shipped and usable.

Desktop shell & Go engine
Live cluster dashboard
Resource browser & inspector
YAML operations & triage

NextMilestone 2

Deep operations

Topology, metrics, networking, storage, and Helm.

Workload topology graph
Prometheus metrics
Port-forward & Cilium
PV/PVC & Longhorn
Helm releases

LaterMilestone 3

Ecosystem

Guardrails, extensibility, and assisted operations.

RBAC & secret guardrails
Plugin SDK
AI-assisted triage
Distribution hardening

Read the full roadmap

FAQ

Questions,
answered plainly.

The docs cover the architecture, security model, and IPC contract in detail. If something is missing, the repository is open.

Read the docs

Bring your kubeconfig. KubeView handles discovery, auth, live watches, and the troubleshooting workbench, all on your machine.

Download v0.1.0

Read the docs

macOS · Windows · Linux / Free for internal use / BSL 1.1

Your cluster,in full view.

Node matrix

One cluster. Three minutes.

Pressure appears.

Read the pod.

Review the fix.

Cluster overview

Built for the pager,not the demo.

Live telemetry, not last-minute polling

Watch streams keep the grid honest

Logs and shell, where the pod is

The whole object, not a summary

Every kubeconfig, one hub

Drift, diffed. Apply, gated.

Command palette first

Your credentials never leave the engine.

Credentials stay in Go memory

stdio, not a TCP port

Agentless by default

Three layers.One clean trust boundary.

Interface

Native host

Kubernetes engine

Cluster

Shipping in the open.

Core IDE

Deep operations

Ecosystem

Questions,answered plainly.

Is KubeView really agentless?

Do I need an account or a cloud backend?

Which clusters does it support?

Where do my credentials live?

Is it free?

Does it replace Argo CD or Flux?

Stop tab-hopping.Start seeing.

Your cluster,in full view.

Node matrix

One cluster. Three minutes.

Pressure appears.

Read the pod.

Review the fix.

Cluster overview

Built for the pager,not the demo.

Live telemetry, not last-minute polling

Watch streams keep the grid honest

Logs and shell, where the pod is

The whole object, not a summary

Every kubeconfig, one hub

Drift, diffed. Apply, gated.

Command palette first

Your credentials never leave the engine.

Credentials stay in Go memory

stdio, not a TCP port

Agentless by default

Three layers.One clean trust boundary.

Interface

Native host

Kubernetes engine

Cluster

Shipping in the open.

Core IDE

Deep operations

Ecosystem

Questions,answered plainly.

Is KubeView really agentless?

Do I need an account or a cloud backend?

Which clusters does it support?

Where do my credentials live?

Is it free?

Does it replace Argo CD or Flux?

Stop tab-hopping.Start seeing.

Three layers.
One clean trust boundary.

Questions,
answered plainly.

Three layers.
One clean trust boundary.

Questions,
answered plainly.