Kubernetes

Kubernetes MCP lets you manage and query Kubernetes resources through a modular MCP server. You can run the MCP in development mode to explore resources, and you can use dedicated MCP tools to fetch, read, and patch resources in your cluster from a local environment.

How to use

You interact with the Kubernetes MCP server by running it in development mode and then using its exposed tools to manage resources. The provided tools let you list available commands, fetch resources by kind and namespace, read specific resources, and patch resources when needed.

To start the MCP server in development mode, use the runtime command shown in the setup flow. Once running, you can use the MCP command set to perform operations against your Kubernetes cluster without exposing raw Kubernetes APIs directly.

How to install

Prerequisites you need before installing the MCP server locally:

• Python and Pip installed on your system
• Goose CLI installed (for tool orchestration)
• Access to a Kubernetes cluster you want to manage

# Step 1: Install Goose CLI (if not already installed)
# This typically uses your system's package manager or a direct install method for Goose.

# Step 2: Install MCP locally
uv pip install .

Configure the Goose CLI to include the Kubernetes MCP extension so you can access the MCP tools from your local shell.

$ goose configure

This will update your existing config file
  if you prefer, you can edit it directly at /home/yurinnick/.config/goose/config.yaml

┌   goose-configure
│
◇  What would you like to configure?
│  Add Extension

◇  What type of extension would you like to add?
│  Command-line Extension

◇  What would you like to call this extension?
│  kube-mcp

◇  What command should be run?
│  uv run /global/path/to/kube-mcp/.venv/bin/mcp-kube

◇  Please set the timeout for this tool (in secs):
│  300

◇  Would you like to add a description?
│  No

◇  Would you like to add environment variables?
│  No

└  Added kube-mcp extension

Validate that the extension is active and the MCP tools are listed.

$ goose

( O)> What tools do you have?
Let me list out the tools that are currently available to me:

1. **Kubernetes MCP Tools**:
   - `kube-mcp__get_resources` - Get resources of a specific kind in a namespace
   - `kube-mcp__get_resource` - Get a specific resource by kind, name, and namespace
   - `kube-mcp__patch_resource` - Patch a specific resource

Let me also check if there are any additional resources available in the kube-mcp extension:
─── list_resources | platform ──────────────────────────
extension_name: kube-mcp

I can also see there are two resources available in the kube-mcp extension:
- `kube://config`
- `kube://context`

Let me check what information these resources provide:
─── read_resource | platform ──────────────────────────
uri: kube://config
─── read_resource | platform ──────────────────────────
uri: kube://context

Additional sections

Development and runtime notes:

• Activate your local virtual environment if you use one, then start the MCP server in development mode using the specified command. This runs the server in a mode suitable for testing and exploration before you deploy to a production environment.

To explore resources, you can use the Kubernetes MCP tools for common operations like listing resources, fetching a specific resource, and patching a resource. These tools are designed to integrate with your local development workflow and provide a streamlined interface for Kubernetes interactions.

Available tools

kube-mcp__get_resources

Fetch resources of a specific kind within a namespace, enabling list-like operations against Kubernetes resources.

kube-mcp__get_resource

Retrieve a single resource by kind, name, and namespace for detail inspection.

kube-mcp__patch_resource

Patch a specific resource to update fields without replacing the entire object.