Kubernetes MCP Server

An MCP server for fetching Kubernetes pod logs. This server provides AI assistants with the ability to list namespaces, pods, and access pod logs.

Features

• List available namespaces
• List pods in a namespace
• Get pod logs (with optional container selection and line limiting)
• Stream pod logs in real-time

Setup

1. Install dependencies:

pip install -e .

2. Configure Kubernetes access:

   Option A: Using Environment Variables (Recommended for specific cluster)

   export K8S_API_SERVER="https://your-k8s-api-server"
   export K8S_TOKEN="your-service-account-token"
   

   Option B: Using Local Kubeconfig

   • The server will automatically try to use in-cluster configuration if running inside Kubernetes
   • Otherwise, it will fall back to using your local kubeconfig file

   For the Mareana dashboard cluster:

   # Use the provided configuration
   source config.example.sh
   
   # Or set manually:
   export K8S_API_SERVER="https://dashboard-dev.mareana.com/api/v1"
   export K8S_TOKEN="test-dummy-token"
   export K8S_VERIFY_SSL="true"
   

   Make sure you have the necessary permissions to access pod logs with your service account.

Running the Server

python -m k8s_mcp_server.server

Available Resources

The server provides the following resources:

1. Namespaces:

   namespace/{namespace_name}
   

2. Pods:

   pod/{namespace_name}/{pod_name}
   

Available Tools

1. Get Pod Logs:

   • Tool: get_pod_logs
   • Parameters:
     • namespace: Kubernetes namespace
     • pod_name: Name of the pod
     • container: (optional) Container name
     • tail_lines: (optional) Number of lines to return from the end

2. List Namespace Pods:

   • Tool: list_namespace_pods
   • Parameters:
     • namespace: Kubernetes namespace