Interaction through cursor

Use Cases

1. Kubernetes Resource Management via LLM

• Interactive Resource Management: Manage Kubernetes resources through natural language interaction with LLM, eliminating the need to memorize complex kubectl commands
• Batch Operations: Describe complex batch operation requirements in natural language, letting LLM translate them into specific resource operations
• Resource Status Queries: Query cluster resource status using natural language and receive easy-to-understand responses

2. Automated Operations Scenarios

• Intelligent Operations Assistant: Serve as an intelligent assistant for operators in daily cluster management tasks
• Problem Diagnosis: Assist in cluster problem diagnosis through natural language problem descriptions
• Configuration Review: Leverage LLM's understanding capabilities to help review and optimize Kubernetes resource configurations

3. Development and Testing Support

• Quick Prototype Validation: Developers can quickly create and validate resource configurations through natural language
• Environment Management: Simplify test environment resource management, quickly create, modify, and clean up test resources
• Configuration Generation: Automatically generate resource configurations that follow best practices based on requirement descriptions

4. Education and Training Scenarios

• Interactive Learning: Newcomers can learn Kubernetes concepts and operations through natural language interaction
• Best Practice Guidance: LLM provides best practice suggestions during resource operations
• Error Explanation: Provide easy-to-understand error explanations and correction suggestions when operations fail

Architecture

1. Project Overview

An stdio-based MCP server that connects to Kubernetes clusters and provides the following capabilities:

• Query Kubernetes resource types (including built-in resources and CRDs)
• CRUD operations on Kubernetes resources (with configurable write operations)
• Helm operations for release and repository management

2. Technical Stack

• Go
• mcp-go SDK
• Kubernetes client-go library
• Helm v3 client library
• Stdio for communication

3. Core Components

1. MCP Server: Uses mcp-go's mcp-k8s package to create an stdio-based MCP server
2. K8s Client: Uses client-go to connect to Kubernetes clusters
3. Helm Client: Uses Helm v3 library for Helm operations
4. Tool Implementations: Implements various MCP tools for different Kubernetes operations

4. Available Tools

Resource Type Query Tools

• get_api_resources: Get all supported API resource types in the cluster

Resource Operation Tools

• get_resource: Get detailed information about a specific resource
• list_resources: List all instances of a resource type
• create_resource: Create new resources (can be disabled)
• update_resource: Update existing resources (can be disabled)
• delete_resource: Delete resources (can be disabled)

Helm Operation Tools

• list_helm_releases: List all Helm releases in the cluster
• get_helm_release: Get detailed information about a specific Helm release
• install_helm_chart: Install a Helm chart (can be disabled)
• upgrade_helm_chart: Upgrade a Helm release (can be disabled)
• uninstall_helm_chart: Uninstall a Helm release (can be disabled)
• list_helm_repositories: List configured Helm repositories
• add_helm_repository: Add a new Helm repository (can be disabled)
• remove_helm_repository: Remove a Helm repository (can be disabled)

Usage

mcp-k8s supports three communication modes:

1. Stdio Mode (Default)

In stdio mode, mcp-k8s communicates with the client through standard input/output streams. This is the default mode and is suitable for most use cases.

# Run in stdio mode (default)
{
    "mcpServers":
    {
        "mcp-k8s":
        {
            "command": "/path/to/mcp-k8s",
            "args":
            [
                "--kubeconfig",
                "/path/to/kubeconfig",
                "--enable-create",
                "--enable-delete",
                "--enable-update",
                "--enable-list",
                "--enable-helm-install",
                "--enable-helm-upgrade"
            ]
        }
    }
}

2. SSE Mode

In SSE (Server-Sent Events) mode, mcp-k8s exposes an HTTP endpoint to mcp client. You can deploy the service on a remote server (but you need to pay attention to security)

# Run in SSE mode
./bin/mcp-k8s --kubeconfig=/path/to/kubeconfig --transport=sse --port=8080 --host=localhost --enable-create --enable-delete --enable-list --enable-update --enable-helm-install
# This command will open all operations

mcp config

{
  "mcpServers": {
    "mcp-k8s": {
      "url": "http://localhost:8080/sse",
      "args": []
    }
  }
}

SSE mode configuration:

• --transport: Set to "sse" to enable SSE mode
• --port: HTTP server port (default: 8080)
• --host: HTTP server host (default: "localhost")

3. Streamable HTTP Mode

In Streamable HTTP mode, mcp-k8s exposes an HTTP endpoint that supports both direct HTTP responses and SSE streams. This mode provides better flexibility and supports streaming output.

# Run in Streamable HTTP mode
./bin/mcp-k8s --kubeconfig=/path/to/kubeconfig --transport=streamable-http --port=8080 --host=localhost --endpoint-path=/mcp --enable-create --enable-delete --enable-list --enable-update --enable-helm-install

mcp config

{
  "mcpServers": {
    "mcp-k8s": {
      "url": "http://localhost:8080/mcp",
      "args": []
    }
  }
}

Streamable HTTP mode configuration:

• --transport: Set to "streamable-http" to enable Streamable HTTP mode
• --port: HTTP server port (default: 8080)
• --host: HTTP server host (default: "localhost")
• --endpoint-path: Endpoint path for the MCP server (default: "/mcp")

4. Docker environment

SSE Mode

1. Complete Example Assuming your image name is mcp-k8s and you need to map ports and set environment parameters, you can run:

docker run --rm -p 8080:8080 -i -v ~/.kube/config:/root/.kube/config ghcr.io/silenceper/mcp-k8s:latest --transport=sse

Streamable HTTP Mode

docker run --rm -p 8080:8080 -i -v ~/.kube/config:/root/.kube/config ghcr.io/silenceper/mcp-k8s:latest --transport=streamable-http --endpoint-path=/mcp

stdio Mode

{
  "mcpServers": {
    "mcp-k8s": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "-v",
        "~/.kube/config:/root/.kube/config",
        "--rm",
        "ghcr.io/silenceper/mcp-k8s:latest"
      ]
    }
  }
}

Getting Started

Direct Usage

You can directly download the binary for your platform from the releases page and use it immediately.

Go Install

go install github.com/silenceper/mcp-k8s/cmd/mcp-k8s@latest

Build

Using Makefile (Recommended)

The Makefile automatically injects version information from the VERSION file:

git clone https://github.com/silenceper/mcp-k8s.git
cd mcp-k8s
make build

This will build with version information injected via ldflags.

Direct Build

git clone https://github.com/silenceper/mcp-k8s.git
cd mcp-k8s
go build -o bin/mcp-k8s cmd/mcp-k8s/main.go

Note: Direct build will show version as "dev" unless you specify ldflags manually.

Command Line Arguments

You can use mcp-k8s --help to see all available options, or mcp-k8s --version to check the version.

Kubernetes Resource Operations

• --kubeconfig: Path to Kubernetes configuration file (uses default config if not specified)
• --enable-create: Enable resource creation operations (default: false)
• --enable-update: Enable resource update operations (default: false)
• --enable-delete: Enable resource deletion operations (default: false)
• --enable-list: Enable resource list operations (default: true)

Helm Operations

• --enable-helm-release-list: Enable Helm release list operations (default: true)
• --enable-helm-release-get: Enable Helm release get operations (default: true)
• --enable-helm-install: Enable Hel