k8s-AIassistant

k8s-AIassistant

A minimal setup to run a Kubernetes MCP server and a UI Plugin backend with Docker Compose.

Overview

• MCP Server (mcp-server-kubernetes)
  • Exposes an HTTP SSE transport (Server-Sent Events) for MCP on port 3000.
  • Talks to your Kubernetes cluster using your kubeconfig.
• UI Plugin Backend (ui-plugin-example/pkg/top-level-product)
  • Node/Express service on port 8055.
  • Connects to an Ollama-compatible OpenAI endpoint and the MCP server.

Prerequisites

• Docker and Docker Compose
• A valid kubeconfig on the host (e.g. /home/<user>/.kube/config)
• An accessible Ollama/OpenAI-compatible endpoint

Quick Start

note: mcp server not support multicluster so you have to chang the KUBECONFIG PATH to

1. Set host user and group for the container to read your kubeconfig:

export UID=$(id -u)
export GID=$(id -g)
chmod 644 /home/<user>/.kube/config

2. Build and start services:

docker compose build
docker compose up -d

3. Verify:

• MCP SSE: curl http://localhost:3000/sse
• UI Plugin: curl http://localhost:8055/health

Configuration

• Defined in docker-compose.yml.
• Service: mcp (port 3000)
  • Environment:
    • ENABLE_UNSAFE_SSE_TRANSPORT=true
    • HOST=0.0.0.0
    • PORT=3000
    • KUBECONFIG_PATH=/kube/config
  • Volume: bind-mount host kubeconfig :/kube/config:ro
  • User: ${UID}:${GID} (set from host) to avoid permission issues
• Service: ui-plugin (port 8055)
  • Environment:
    • MCP_BASE=http://mcp:3000 (service name inside the compose network)
    • `OLLAMA_BASE=
    • MODEL_NAME=gpt-oss:20b

Adjust OLLAMA_BASE, MODEL_NAME, kubeconfig path, and ports to match your environment.

Demo

• Screenshot demo:

Project Layout

• MCP server source: mcp-server-kubernetes/
• UI plugin backend: ui-plugin-example/pkg/top-level-product/pages/server.js
• UI plugin Dockerfile: ui-plugin-example/pkg/top-level-product/Dockerfile