k8s_mcp_server_prod

This MCP server provides a natural-language interface for interacting with a Kubernetes cluster. It uses a FastAPI backend that serves an MCP schema describing the supported commands, and it leverages Google's Gemini model via kubectl-ai to translate natural language prompts into Kubernetes operations. Once the MCP server is running, you can connect your kubectl-ai client to the endpoint and start issuing high-level prompts like listing pods, scaling deployments, or inspecting cluster resources. The server exposes an MCP schema JSON at /mcp-schema.json which clients can fetch to understand the available commands and their required parameters. To use it in a workflow, configure kubectl-ai with the endpoint and your Gemini API credentials, then issue prompts such as: “List all pods in the default namespace”, “Scale my-website-app to 5 replicas”, or “Get the status of all nodes.” The integration abstracts Kubernetes complexity behind concise, NL prompts while preserving the ability to perform precise actions when needed.

Prerequisites:

• Ubuntu 22.04+ or compatible Linux environment
• Python 3.9+ and pip
• Access to a Kubernetes cluster (local Minikube or remote) and kubectl installed
• Gemini API access key

Install steps:

1. Prepare environment

sudo apt-get update
sudo apt-get install -y python3 python3-pip git

2. Clone the MCP server repository

git clone https://github.com/your-organization/k8s-mcp-server-prod.git
cd k8s-mcp-server-prod

3. Install Python dependencies

pip3 install -r requirements.txt

4. Configure environment variables

• Create a config file or export variables:

export GEMINI_API_KEY=your-key
export GEMINI_API_ENDPOINT=https://gemini.googleapis.com/v1
export KUBECONFIG=~/.kube/config

5. Run the MCP server

python -m uvicorn main:app --host 0.0.0.0 --port 8000

6. (Optional) Run inside Docker or Kubernetes for production

• Build a container with the Python runtime and uvicorn startup, then deploy to your cluster or container platform.
• Expose the service via a NodePort or LoadBalancer, and ensure the MCP schema endpoint is reachable at http://<host>:8000/mcp-schema.json

Tips and caveats:

• The MCP schema at /mcp-schema.json defines the supported prompts and their required parameters; fetch it before crafting prompts to ensure valid commands.
• Ensure the Gemini API key has the required permissions and that the endpoint is accessible from your deployment environment.
• If you’re running in Kubernetes, consider mounting the kubeconfig as a secret and referencing it via an environment variable or volume.
• For production, run uvicorn with a process manager (e.g., gunicorn) and behind an HTTPS ingress to protect API traffic.
• When prompts fail or return unexpected Kubernetes actions, enable verbose logs in FastAPI to diagnose translation failures from NL prompts to Kubernetes commands.