zane — Chat with your Kubernetes cluster

what makes it different

Not a command.
A conversation.

💬

Ask anything in plain English

No flags, no man pages. Just ask "why is my api pod crashing?" and Zane figures out what to fetch, reads the evidence, and answers with citations.

conversational

🔍

Evidence-cited answers

Every answer references the exact log line, event, or resource field that led to it. No hallucinations — if Zane says it, it came from your cluster.

trustworthy

🛡️

Write operations need your OK

Zane can apply fixes, but never silently. Every write — patch, delete, restart — shows you the exact command and waits for y/N before touching anything.

safe by default

🧠

Session history

Zane remembers your conversation. Ask follow-up questions, drill into a specific pod, or revisit an earlier finding — all within the same session.

stateful

🧙

First-run wizard

Run `zane` for the first time and it walks you through API key setup, kubeconfig, telemetry opt-in, and session history. Zero config files to hand-edit.

zero friction

📊

Feedback built in

Type /good or /bad after any answer. Zane improves from real operator feedback — not synthetic benchmarks. Your signals shape what gets better next.

/good · /bad · /clear

how it works

Ask. Inspect. Fix.
In one session.

01 ── start

Just run zane

First run launches a setup wizard — API key, kubeconfig, telemetry. After that it's instant. Zane connects to whichever cluster your current context points at.

$ zane
✦ Connected to production (3 nodes)
✦ What would you like to know?

02 ── ask

Type your question

Ask anything a senior SRE would ask. Zane's agent loop picks the right tools, fetches the relevant data from your cluster, and cites the exact source in its answer.

you: why is api-server crashing?
→ fetching logs, events, limits
zane: OOM Kill — 512Mi limit hit

03 ── fix

Apply with confirmation

Ask Zane to apply the fix. It shows you the exact kubectl command it will run and waits for your confirmation before touching anything in the cluster.

zane: Patch memory limit to 1Gi?
kubectl patch deploy/api ...
Apply? [y/N]: y ✓ done

built-in tools

14 cluster tools.
12 read. 2 write.

Zane picks the right tool automatically based on your question.

📋 get pods

📜 get logs

🔍 describe pod

📡 get events

🗂️ get deployments

⚙️ get configmaps

🔒 get secrets (redacted)

💾 get nodes

📦 get namespaces

🔗 get services

📊 top pods

📈 top nodes

✏️ apply patch

🔄 restart pod

▌ read — always safe, no confirmation needed ▌ write — always requires y/N confirmation

vision

Built for the full
Kubernetes operator stack

Kubernetes operators don't just run kubectl. They manage ArgoCD deployments, debug Helm release diffs, investigate Prometheus alerts, trace Grafana anomalies — then switch back to pods.

Zane is being built to be the single conversational layer across all of it. The more of your stack you bring, the more Zane understands.

☸️

Kubernetes

Pods, deployments, nodes, events, logs

live today

🐙

ArgoCD

App health, sync status, rollback reasoning

roadmap

⎈

Helm

Release diffs, failed upgrades, value drift

roadmap

🔥

Prometheus

Alert context linked back to cluster state

roadmap

📊

Grafana

Anomaly → pod correlation, dashboard context

roadmap

🕸️

Istio / Linkerd

Traffic errors, mTLS failures, mesh topology

roadmap

🔐

cert-manager

Expiring certs, ACME failures, issuer issues

roadmap

🪣

Velero

Backup failures, restore status, schedule drift

roadmap

● works today

▸Pods, deployments, nodes, namespaces

▸Logs, events, resource limits

▸CrashLoopBackOff root cause

▸Pending pod scheduling failures

▸Stuck deployment rollouts

▸Write ops with y/N confirmation

▸Session history + feedback commands

◌ on the roadmap

▸ArgoCD app health + sync reasoning

▸Helm release diffs + failed upgrades

▸Prometheus alert → cluster linkage

▸Grafana anomaly correlation

▸Istio / Linkerd traffic errors

▸cert-manager certificate issues

▸RAG on past incidents (pattern memory)

The goal: one conversation to understand any layer of your Kubernetes platform.

install

One binary.
Zero cluster changes.

# requires Go 1.21+
go install github.com/zarakm/zanecli@latest

# run — wizard handles the rest
zanecli

# macOS — Apple Silicon (M1/M2/M3)
curl -L -o zanecli.tar.gz https://github.com/zarakM/zanecli/releases/download/v0.1.0/zanecli_0.1.0_Darwin_arm64.tar.gz
tar -xzf zanecli.tar.gz zanecli
sudo mv zanecli /usr/local/bin/
xattr -d com.apple.quarantine /usr/local/bin/zanecli 2>/dev/null
zanecli

The xattr line clears the macOS Gatekeeper "unsigned binary" block.

# macOS — Intel
curl -L -o zanecli.tar.gz https://github.com/zarakM/zanecli/releases/download/v0.1.0/zanecli_0.1.0_Darwin_x86_64.tar.gz
tar -xzf zanecli.tar.gz zanecli
sudo mv zanecli /usr/local/bin/
xattr -d com.apple.quarantine /usr/local/bin/zanecli 2>/dev/null
zanecli

The xattr line clears the macOS Gatekeeper "unsigned binary" block.

# Linux — x86_64
curl -L -o zanecli.tar.gz https://github.com/zarakM/zanecli/releases/download/v0.1.0/zanecli_0.1.0_Linux_x86_64.tar.gz
tar -xzf zanecli.tar.gz zanecli
sudo mv zanecli /usr/local/bin/
zanecli

# Linux — arm64
curl -L -o zanecli.tar.gz https://github.com/zarakM/zanecli/releases/download/v0.1.0/zanecli_0.1.0_Linux_arm64.tar.gz
tar -xzf zanecli.tar.gz zanecli
sudo mv zanecli /usr/local/bin/
zanecli

# Windows — x86_64 (PowerShell)
Invoke-WebRequest -Uri https://github.com/zarakM/zanecli/releases/download/v0.1.0/zanecli_0.1.0_Windows_x86_64.zip -OutFile zanecli.zip
Expand-Archive zanecli.zip -DestinationPath .
.\zanecli.exe

# via kubectl plugin manager
kubectl krew install zane

zanecli

Requires Krew. Submission to Krew index in progress.

✦ First run walks you through setup automatically — API key, kubeconfig, telemetry opt-in. No config files to hand-edit.

A conversational AI co-pilotfor Kubernetes operators

A conversational AI co-pilot
for Kubernetes operators