Cyclops

Cyclops MCP lets your AI agents safely manage Kubernetes applications by operating at the level of Cyclops Modules, abstracting away low-level Kubernetes resources. It enables rapid, production-grade changes with reduced risk by guiding the agent through high-level resources.

How to use

You connect your MCP client to the Cyclops MCP server and start new conversations with your AI companion. The server exposes a stream you can subscribe to and accepts module-based actions that translate into Kubernetes deployments, services, and more as high-level Cyclops Modules. Use this to let your AI agent propose, validate, and apply changes without you needing to micro-manage every Kubernetes detail.

How to install

Prerequisites: ensure you have a Kubernetes cluster and kubectl configured on your workstation.

Install via UI (recommended) Start Cyclops MCP in the UI by selecting Addons → MCP server and clicking Install Cyclops MCP server. Use a Cyclops version of at least v0.20.1. Expose the MCP server outside the cluster by port-forwarding. Run:

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

Your server will be reachable at localhost:8000.

Connect to an AI agent by providing the MCP server address in its configuration. Here is an example you can use with a client that accepts MCP server definitions:

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

## Install stdin binary

Install the MCP server binary for local use with stdin transport.

1. Make sure Cyclops is installed in your Kubernetes cluster. You generally install it with a manifest in your cluster. (Follow your local workflow to install Cyclops and required CRDs.)

2. Download MCP server binary

GOBIN="$HOME/go/bin" go install github.com/cyclops-ui/mcp-cyclops/cmd/mcp-cyclops@latest

3. Add server configuration

{ "mcpServers": { "mcp-cyclops": { "command": "mcp-cyclops" } } }

## Configuration and environment

You can configure Cyclops MCP to specify which kubeconfig file to use and other Kubernetes context options. The following environment variables are commonly provided to customize behavior.

{ "mcpServers": { "mcp-cyclops": { "command": "mcp-cyclops", "env": { "KUBECONFIG": "/path/to/your/kubeconfig" } } } }

## Manual cluster installation

If you install Cyclops MCP directly into a Kubernetes cluster and want to connect multiple developers, you can expose the MCP server via a service and port-forward for testing.

1. Ensure Cyclops components and CRDs are installed in the cluster.

kubectl get pods -n cyclops

2. Check that CRDs are installed.

kubectl get crds | grep cyclops-ui

3. Apply the MCP server manifest to install the server.

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

4. Expose the MCP server and test with port-forward.

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

5. Add the MCP server to your client configuration.

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

## Available tools

### create\_module

Create a new Module after validating its values with get\_template\_schema.

### get\_module

Fetch a Module by name.

### list\_modules

List all Cyclops Modules.

### update\_module

Update a Module by name after validating its values with get\_template\_schema.

### get\_template\_schema

Return the JSON schema for a given template to validate values before creating a Module.

### get\_template\_store

Fetch a Template Store by name.

### list\_template\_store

List Template Stores available in the cluster.