Kopilot MCP

What is Kopilot MCP?

Kopilot MCP is an AI‑powered Kubernetes agent built on the Model Context Protocol (MCP). It runs as a command‑line application for Python 3.12+ and lets developers manage Kubernetes resources by typing natural language queries instead of kubectl commands.

How to use Kopilot MCP?

Clone the repository, copy config/dev/example.toml to config/dev/config.toml and edit it, then install dependencies with uv sync. Run uv run mcp_client.py to start the client (it launches the MCP server automatically). At the Query prompt, type natural‑language requests such as “Get po in kube‑system ns” to list pods.

Key features of Kopilot MCP

• Creates an AI Kubernetes agent using MCP.
• Interact with Kubernetes resources using natural language.
• Customize LLM provider for “kubectl‑like” commands.
• Learn how MCP standardizes tool calling.

Use cases of Kopilot MCP

• List pods in a namespace with a plain‑English query.
• Retrieve logs or describe Kubernetes objects without memorizing commands.
• Experiment with MCP tool‑calling patterns for Kubernetes automation.
• Quickly debug a cluster by asking about running resources.

FAQ from Kopilot MCP

What prerequisites do I need?

Python 3.12+, the uv package manager, an API key from an LLM provider (e.g., OpenRouter), and access to a Kubernetes cluster (such as a local kind cluster).

How do I configure the LLM provider?

Edit the config/dev/config.toml file (rename from the provided example.toml) to set your API key and model preferences. The project uses LangChain and OpenAI but works with any provider that offers an OpenAI‑compatible endpoint.

Does Kopilot MCP connect to any Kubernetes cluster?

Yes. It uses the official kubernetes Python client, so it works with any cluster you have configured in your kubeconfig (e.g., kind, Minikube, or a remote cluster).

Is Kopilot MCP free and open‑source?

Yes. It is released under the MIT License, so you can freely use, modify, and distribute it.