mcp-cyclops

Cyclops MCP provides an SSE-based server that your AI agents can talk to in order to manage Kubernetes resources through high-level Cyclops Modules rather than low-level Kubernetes objects. The server exposes a set of tools for creating, updating, listing, and validating Modules and Templates, and it enforces templates and schemas to help prevent misconfigurations in production. Tools include create_module, update_module, get_module, list_modules, get_template_schema, get_template_store, and list_template_store, allowing you to generate, inspect, and apply modular deployments with guidance from predefined schemas. To connect an AI agent, run the MCP server in your cluster or on your preferred host, expose it via SSE, and point the agent's configuration at the server URL (for example, http://localhost:8000/sse when port-forwarded). The server is designed to simplify Kubernetes complexity by focusing on Cyclops Modules as the unit of work, thereby reducing opportunities for misconfigurations during automated planning and execution.

Prerequisites:

• Access to a Kubernetes cluster with kubectl configured
• Go installed if you plan to install the binary via go install
• Optional: Cyclops UI version v0.20.1 or greater if installing through the UI

Install via binary (recommended when not using UI):

1. Install the Cyclops MCP server binary:

# Install the mcp-cyclops binary (Go must be available)
GOBIN="$HOME/go/bin" go install github.com/cyclops-ui/mcp-cyclops/cmd/mcp-cyclops@latest

2. Run or configure the server via mcp-config (see below) and ensure it can reach your Kubernetes cluster.

3. Expose the MCP server as needed (e.g., port-forward for local testing):

kubectl port-forward svc/cyclops-mcp -n cyclops 8000:8000

4. Connect your MCP server in your agent configuration, providing the server URL (for SSE) as shown in the examples.

Install via Kubernetes manifest (manual cluster install):

1. Ensure Cyclops components and CRDs are installed in your cluster as described in the Cyclops docs.
2. Apply the MCP server manifest:

kubectl apply -f https://raw.githubusercontent.com/cyclops-ui/mcp-cyclops/refs/heads/main/install/mcp-server.yaml

3. Expose the service if needed (port-forward example):

kubectl port-forward svc/cyclops-mcp -n cyclops 8000:8000

4. Add your MCP server to the agent configuration (SSE URL):

{
  "mcpServers": {
    "mcp-cyclops": {
      "url": "http://localhost:8000/sse"
    }
  }
}

Environment variables: The server can be configured with a kubeconfig path and various Cyclops-specific settings. Common vars include KUBECONFIG (path to kubeconfig), CYCLOPS_KUBE_CONTEXT (Kubernetes context), and several CYCLOPS_ prefixed namespaces for modules and Helm releases. If you run inside a cluster, you can omit KUBECONFIG and rely on in-cluster configuration. When exposing the server for external agents, ensure proper security (TLS/auth) for SSE endpoints. If you encounter port-forward or networking issues, verify the Cyclops module namespaces and that the cyclops-mcp service is reachable from the agent environment.