K8s Cluster Ops MCP Server

Manage Kubernetes clusters through natural language — list resources, inspect logs, scale deployments, apply manifests, and monitor cluster health. All mutating operations require explicit confirmation.

Quick Start

1. Deploy on Apify — search "K8s Cluster Ops MCP" in the Apify Store, or run locally with Docker.
2. Add the MCP server URL to your Claude Desktop config (claude_desktop_config.json):

   {
     "mcpServers": {
       "k8s-cluster-ops": {
         "url": "https://your-apify-actor-standby-url/sse"
       }
     }
   }
   

3. Start managing your cluster with natural language in Claude.

Run locally

git clone <repo-url>
cd products/mcp-servers/k8s-cluster-ops

# Ensure kubectl is installed and kubeconfig is set
export KUBECONFIG="$HOME/.kube/config"

pip install -r requirements.txt
python src/server.py

The server starts on port 8080 by default (configure with PORT env var).

Configuration

Parameter	Required	Description	Default
KUBECONFIG	No	Path to kubeconfig file	~/.kube/config
KUBECTL_PATH	No	Path to kubectl binary	kubectl
HOST	No	Server bind address	0.0.0.0
PORT	No	Server port	8080

Available Tools

list_namespaces

List all namespaces in the cluster with their status and labels.

Parameters: None.

Example prompt: "What namespaces exist in my cluster?"

Example response:

{
  "count": 4,
  "namespaces": [
    { "name": "default", "status": "Active", "labels": {} },
    { "name": "production", "status": "Active", "labels": { "env": "prod" } },
    { "name": "staging", "status": "Active", "labels": { "env": "staging" } },
    { "name": "kube-system", "status": "Active", "labels": {} }
  ]
}

list_pods

List all pods in a namespace with status, readiness, and restart counts.

Parameters: namespace (required).

Example prompt: "Show me all pods in the production namespace"

list_deployments

List deployments in a namespace with replica counts and images.

Parameters: namespace (required).

Example prompt: "What deployments are running in staging?"

list_services

List services in a namespace with type, ClusterIP, and port mappings.

Parameters: namespace (required).

Example prompt: "List all services in the production namespace"

get_pod_logs

Tail logs from a specific pod.

Parameters: namespace (required), pod_name (required), container (optional), tail_lines (optional, default 100, max 1000).

Example prompt: "Show me the last 50 lines of logs from the api-server pod in production"

describe_resource

Get a detailed description of any Kubernetes resource (equivalent to kubectl describe).

Parameters: namespace (required), resource_type (required), resource_name (required).

Example prompt: "Describe the redis deployment in the cache namespace"

scale_deployment

Scale a deployment's replica count. Requires confirm: true.

Parameters: namespace (required), deployment_name (required), replicas (required), confirm (required, must be true).

Example prompt: "Scale the web-frontend deployment in production to 5 replicas"

Without confirm: true, returns a preview:

{
  "warning": true,
  "message": "This will scale deployment 'web-frontend' in namespace 'production' to 5 replicas. To proceed, call this tool again with confirm=true.",
  "confirm_required": true
}

apply_manifest

Apply a YAML manifest to the cluster. Supports dry-run mode. Requires confirm: true for actual apply.

Parameters: manifest_yaml (required), namespace (optional), dry_run (optional, default false), confirm (required for non-dry-run).

Example prompt: "Apply this ConfigMap to the staging namespace (dry run first)"

get_cluster_health

Aggregate cluster health status — nodes, PVCs, and namespace overview.

Parameters: None.

Example prompt: "How healthy is my cluster right now?"

restart_deployment

Perform a rolling restart of a deployment. Requires confirm: true.

Parameters: namespace (required), deployment_name (required), confirm (required, must be true).

Example prompt: "Rolling restart the api-server deployment in production"

Safety Guardrails

• Mutating operations (scale_deployment, apply_manifest, restart_deployment) require confirm: true. Without it, the tool returns a warning describing what would happen.
• Blocked operations: kubectl delete namespace and kubectl delete node are permanently blocked.
• Dry-run support: apply_manifest supports dry_run: true to preview changes without applying.
• Namespace required: All pod/deployment/service operations require an explicit namespace — the server never defaults to default.

Example Prompts

• "What namespaces exist in my cluster?"
• "Show me all pods in the production namespace"
• "Are there any pods crashing in staging?"
• "Show me the logs for the auth-service pod in production"
• "Scale the web-frontend to 3 replicas in staging"
• "How healthy is my cluster?"
• "Apply this deployment manifest to production (dry run first)"
• "Rolling restart the api-gateway in production"

Pricing

Plan	Price	Includes
Starter	$49/mo	Read-only tools (list, logs, describe, health)
Pro	$149/mo	All tools including scale, apply, and restart

Purchase on Apify Store or MCPize.

Local Development

# Install dependencies
pip install -r requirements.txt

# Run with hot reload (development)
uvicorn src.server:app --reload --port 8080

# Run tests
pytest tests/ -v

# Build Docker image
docker build -t k8s-cluster-ops-mcp .
docker run -p 8080:8080 \
  -v "$HOME/.kube/config:/root/.kube/config:ro" \
  k8s-cluster-ops-mcp

Troubleshooting

Issue	Solution
kubectl not found	Install kubectl or set KUBECTL_PATH to its location
connection refused	Ensure your kubeconfig points to a reachable cluster
forbidden	Check RBAC — your kubeconfig user needs appropriate permissions
confirm required	Mutating operations need confirm: true — this is by design
Timeout errors	The cluster may be unreachable; check network connectivity