mahradbt

Kubescout MCP — DevOps skill for Claude Code

DevOps community

Read-only by default MCP server for diagnosing Kubernetes incidents.

How to install Kubescout MCP

This entry records only its repository, not the path inside it, so there is no exact command to give. Open mahradbt/kubescout-mcp and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Kubescout MCP does

Read-only by default MCP server for diagnosing Kubernetes incidents. Gives Claude Code and other AI agents one structured summary of pod status, events, logs and recent rollouts, with an explicit allowlist for any write action.

Alternatives in DevOps

Kubeshark — Cluster-wide network observability for Kubernetes 11.8k ★
Mirrord — Run any process, on your machine or in an AI agent's environment, as if it were a pod in your Kubernetes clust 5.3k ★
MCP Boilerplate — A remote Cloudflare MCP server boilerplate with user authentication and Stripe for paid tools 1k ★

README

kubescout-mcp

A read-only Kubernetes MCP server for Claude Code, built with FastMCP in Python.

Quick start

Install:

git clone  kubescout-mcp && cd kubescout-mcp
python3 -m venv .venv
.venv/bin/pip install -e .

Add it to Claude Code:

claude mcp add --scope user kubescout-mcp -- $(pwd)/.venv/bin/kubescout-mcp
claude mcp list

The list should show `kubescout-mcp ... Connected`. Restart your Claude Code session, then try:

List all pods in every namespace and tell me which ones are not ready.

Why this exists

Checking a cluster from an AI assistant usually means pasting `kubectl` output by hand. This server gives Claude Code direct, read-only access to your cluster, so it can look up pods, logs and events by itself. It cannot change anything, so it is safe to point at a real cluster.

Features

Lists namespaces, nodes, pods, deployments and services.
Reads pod logs, with an optional container and a line limit (1 to 1000).
Shows recent namespace events to help debug failing pods.
Returns compact JSON (readiness, restarts, replicas, images) instead of full API objects.
Uses your current kubeconfig context, and falls back to in-cluster config.
Includes a small demo cluster setup (k3d, SQLite and nginx) to try the tools on.

Tools

Tool	Purpose	Mode
`list_namespaces`	All namespaces with status and creation time	read-only
`list_nodes`	Nodes with readiness, kubelet version, OS image, CPU and memory	read-only
`list_pods`	Pods with phase, ready count, restarts, node and IP	read-only
`list_deployments`	Deployments with desired, ready and available replicas, and images	read-only
`list_services`	Services with type, cluster IP and ports	read-only
`get_pod_logs`	Last N