What is Kopilot Mcp

Kopilot-MCP is a remake of a previous project that allows users to create an AI Kubernetes agent using the Model Context Protocol (MCP).

Use cases

Use cases for Kopilot-MCP include querying Kubernetes resources, automating deployment processes, and simplifying the management of Kubernetes clusters through conversational interfaces.

How to use

To use Kopilot-MCP, clone the repository, configure the necessary files, install dependencies, and run the MCP client to interact with Kubernetes resources using natural language.

Key features

Key features include the ability to create an AI Kubernetes agent, interact with Kubernetes resources using natural language, customize LLM providers for ‘kubectl-like’ commands, and learn to standardize tool calling with MCP.

Where to use

Kopilot-MCP can be used in DevOps environments, cloud infrastructure management, and any scenario where Kubernetes resources need to be managed efficiently through natural language.

Kopilot MCP

A MCP remake 🔧 of my previous project.

Key Features

• Create an AI Kubernetes agent using MCP.
• Interact with Kubernetes resources using natural language.
• Customize LLM provider to perform “kubectl-like” commands.
• Learn how to use MCP to standardize tool calling.

Tech Stack

• Dynaconf
• kubernetes
• LangChain
• Loguru
• OpenAI

Getting Started

Prerequisites

• Python 3.12+
• uv
• A API key from LLM provider (e.g. OpenRouter)
• A Kubernetes cluster. (e.g. kind)

Setup

1. Clone this repository.
2. Configure config/dev/example.toml and rename it to config/dev/config.toml.
3. Install dependencies.

uv sync

Run

1. Run the MCP client. It will automatically start the MCP server.

uv run mcp_client.py

2. Then you would see a prompt in the terminal:

Query (Type `exit`, `quit`, `q` to quit):

3. Interact with natural language to operate Kubernetes resources. For example:

Query (Type `exit`, `quit`, `q` to quit): Get po in kube-system ns

4. You would get the response similar like:

Here are the pods in the `kube-system` namespace:

- coredns-7db6d8ff4d-lz4s8
- coredns-7db6d8ff4d-qbgvc
- etcd-test-control-plane
- kindnet-cnjm6
- kube-apiserver-test-control-plane
- kube-controller-manager-test-control-plane
- kube-proxy-v647r
- kube-scheduler-test-control-plane
- sealed-secrets-controller-67767c668-dz4bj

Project Structure

kopilot-mcp
├── README.md             # Project documentation
├── config                # Configuration
│   ├── config.py
│   └── dev
│       └── example.toml
├── mcp_client.py         # MCP client
├── mcp_server.py         # MCP server
├── mcp_server_factory.py # MCP server factory
├── pyproject.toml        # Python project meta
├── scheme                # scheme (Kubernetes GVR ➡ GVK)
│   ├── __init__.py
│   └── scheme.py
├── utils                 # Utilities
│   ├── __init__.py
│   ├── clients.py
│   ├── helpers.py
│   └── models.py
└── uv.lock               # uv lock file

Workflow

Contributing

1. Fork the repository.
2. Create your feature branch.
3. Submit a pull request.

License

MIT License - See LICENSE file for details.