mcpField
K
kube-noob

kube-noob

Local
github.comCloud & Infra0Updated today

About kube-noob

kube-noob provides Kubernetes, OpenShift, ROSA HCP, Helm and RHACS auditing, debugging, documentation and exercises without embedding an AI model. It is for operators and learners using an MCP client or CLI, with read-only-first playbooks and cited local knowledge.

Key features

  1. 01Audit rules and hardened manifests
  2. 02Offline debugging playbooks
  3. 03Cited local documentation
  4. 04Graded learning tracks and exercises
  5. 05Read-only cluster evidence, excluding Secrets
  6. 06CLI operation without an AI model

Use cases

  1. 01Audit Kubernetes or OpenShift manifests
  2. 02Investigate errors with offline playbooks
  3. 03Explain policy denials and network reachability
  4. 04Learn platform operations through graded exercises

MCP support

Exposes audit, debugging and related tools to MCP clients, including Claude and OpenShift Lightspeed. Cluster and web tools are available when enabled.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"aashishmahajan-kube-noob": {
"command": "uvx",
"args": [
"kube-noob"
],
"env": {
"KUBE_NOOB_WEB_ENABLED": "<KUBE_NOOB_WEB_ENABLED>",
"KUBE_NOOB_CLUSTER_ENABLED": "<KUBE_NOOB_CLUSTER_ENABLED>",
"KUBE_NOOB_OFFLINE": "<KUBE_NOOB_OFFLINE>"
}
}
}
}
Run commanduvx kube-noob
Environment variables to set
KUBE_NOOB_WEB_ENABLEDAllow allowlisted web lookups (default false: offline)
KUBE_NOOB_CLUSTER_ENABLEDEnable read-only live cluster tools using your kubectl/oc context (default false)
KUBE_NOOB_OFFLINEAdmin lock: no web, no downloads, local models only (default false)

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypikube-noob0.6.0

Tools

31

Saved tool list · updated today

ParameterTypeRequiredDescription
requeststringYesWhat the user wants, in their words.
versionanyNo
platformanyNo
blueprintanyNoGenerator blueprint (see generate_manifests).
manifestsanyNoExisting manifests to review.
apply_fixesbooleanNo

Details

From the project README

kube-noob - No AI inside. Works with yours.

kube-noob

ci OpenSSF Scorecard PyPI License

No AI inside. Works with yours. kube-noob is a Kubernetes, OpenShift, ROSA HCP, Helm and RHACS/StackRox expert in a box: audit rules, read-only-first playbooks, cited docs and graded exercises. Claude, OpenShift Lightspeed or any MCP client calls it as tools; the CLI runs it with no AI at all. It makes no network calls unless you turn them on.

  • Audit: 60 rules plus strict schemas (Kubernetes 1.35, Routes, SCCs, EgressFirewall, Gateway API, StackRox), removed-API checks, safe auto-fixes, 9 red-team scenarios, RHACS policy mapping.
  • Generate: hardened manifests and Helm charts, re-audited before return.
  • Debug: 38 offline playbooks, read-only steps first; reported OpenShift bugs with their fix version; policy denials explained (Kyverno, Gatekeeper, Pod Security, SCC, RBAC); NetworkPolicy reachability; read-only cluster evidence (never Secrets).
  • Learn: 5 tracks, 16 lessons, graded exercises; remembers confirmed fixes locally.
  • Cite: answers cite local knowledge as [kn:<doc>#<n>].

Every design request runs a 7-step protocol: Clarify, Validate, Research, Audit, Calibrate, Generate, Red Team.

Who does the thinking

You useWho answersWhat leaves your machine
The Claude plugin (Claude Code, Cowork)Claude, calling kube-noob's toolsWhat you share with Claude, as in any chat; kube-noob sends nothing
OpenShift Lightspeed or another MCP clientThat client's modelWhatever that client sends its model; kube-noob sends nothing
The CLIkube-noob's rules, no modelNothing
kube-noob ask (optional)A model you configure: Anthropic or a local oneYour question, to that model only

Quick start

brew tap aashishmahajan/kube-noob https://github.com/aashishmahajan/kube-noob
brew trust --formula aashishmahajan/kube-noob/kube-noob
brew install kube-noob
kube-noob demo --list
kube-noob audit -f k8s/ --platform openshift --version 4.22
kube-noob diagnose 'paste events or errors'
kube-noob doctor
kube-noob selftest

Short form: noob. No Homebrew: uv tool install kube-noob or pipx install kube-noob (Python 3.11+, not macOS's built-in 3.9). Errors: install. Roles, demos: use cases.

Use with Claude or any MCP client

Claude plugin, in Claude Code (needs uv):

/plugin marketplace add aashishmahajan/kube-noob
/plugin install kube-noob@kube-noob

In claude.ai or the desktop app: Customize > Plugins > Add > Add marketplace, aashishmahajan/kube-noob.

Any MCP client:

{"mcpServers": {"kube-noob": {"command": "uvx", "args": ["--from", "kube-noob==0.4.0", "kube-noob", "mcp"]}}}

Cluster and web tools appear when enabled.

Run on OpenShift

As a token-protected service, a loopback-only sidecar, or a tool for OpenShift Lightspeed: docs/openshift.md. The chart ships NetworkPolicies and runs under restricted-v2.

Safe by default

SettingDefaultChange with
Web lookups (HTTPS GET, allowlisted docs, text only)offwebEnabled: true / KUBE_NOOB_WEB_ENABLED=true
Admin offline lock (overrides web, local models only)offoffline: true / KUBE_NOOB_OFFLINE=1 / chart config.offline
Live cluster (read-only, no Secrets)offcluster.enabled: true / --cluster-enabled
Local learning file (profile, confirmed notes)onlearningEnabled: false
Model for kube-noob ask onlynonellm.provider: anthropic (kube-noob[ask]) or openai-compatible

kube-noob doctor shows what it can reach. Config never comes from the current directory implicitly. All keys: kube-noob.example.yaml.

Docs

Credits

Credited sources: THIRD_PARTY_NOTICES.md. Kubernetes, OpenShift, Helm, ROSA, RHACS and StackRox are trademarks of their owners; kube-noob is independent and not endorsed by them. Licensed under Apache-2.0.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.