K8s Pilot

You manage and observe multiple Kubernetes clusters from a single centralized control plane. This MCP server lets you perform common Kubernetes operations across fleets, switch contexts between clusters, and inspect resources safely in readonly mode. It’s designed to streamline multi-cluster administration with a focused, command-driven interface.

How to use

Connect to your clusters and perform standard operations through an MCP client. You can switch between clusters, list and view resources, and create or modify deployments, services, and other Kubernetes objects. When you enable readonly mode, you can inspect resources without risking any changes.

How to install

Prerequisites you need before running the server:

• Python 3.13 or higher

• uv package manager

• Access to your Kubernetes clusters (via ~/.kube/config or in-cluster config)

Follow these steps to install and run the MCP server locally.

Additional sections

Usage in normal and readonly modes is shown below. In normal mode you have full read/write access to manage your clusters. In readonly mode, write operations are blocked to keep changes safe while you inspect resources.

Actual server run configurations

uv run --with mcp[cli] mcp run k8s_pilot.py

For readonly mode run this variant in your environment.

uv run --with mcp[cli] python k8s_pilot.py --readonly

Claude Desktop configurations for running the MCP server

Use these MCP server configurations within Claude Desktop to launch the k8s_pilot server. Use the placeholder path when you clone the repository.

{
  "mcpServers": {
    "k8s_pilot": {
      "command": "uv",
      "args": [
        "--directory",
        "<path-to-cloned-repo>/k8s-pilot",
        "run",
        "--with",
        "mcp[cli]",
        "mcp",
        "run",
        "k8s_pilot.py"
      ]
    }
  }
}

For readonly mode, use this configuration.

{
  "mcpServers": {
    "k8s_pilot_readonly": {
      "command": "uv",
      "args": [
        "--directory",
        "<path-to-cloned-repo>/k8s-pilot",
        "run",
        "--with",
        "mcp[cli]",
        "python",
        "k8s_pilot.py",
        "--readonly"
      ]
    }
  }
}

Available tools

multi_cluster_context_switching

Switch context between multiple Kubernetes clusters and perform operations in the selected cluster.

resource_crud

Create, read, update, and delete common Kubernetes resources such as Deployments, Services, Pods, ConfigMaps, Secrets, Ingresses, StatefulSets, DaemonSets, Roles, and PersistentVolumes/Claims.

readonly_mode

Enable a readonly safety mode that prevents any write operations while allowing read operations for inspection and auditing.

claude_integration

Powered by MCP for Claude AI and extended capabilities within the MCP ecosystem.