Content
<p align="center" width="100%">
<img width="75%" src="https://raw.githubusercontent.com/cyclops-ui/cyclops/main/web/static/img/cyclops-simplistic.png">
<p/>
# Cyclops MCP (Model Context Protocol)
Cyclops MCP allows your favorite AI agent to manage your Kubernetes applications. Cyclops MCP servers provide tools for agents to create and update existing applications safely.
https://github.com/user-attachments/assets/0c675c33-1e36-4fdb-bf8c-2fd7fb491e6b
This means it can check all of your existing templates and the schema of those templates to create accurate and production-ready applications. Your agent now has much less room to make a misconfiguration since it creates high-level resources (Cyclops Modules) instead of touching every line of your Kubernetes resources (Deployments, Services, and Ingresses).
It allows you to move fast and ensure no uncaught misconfigurations are hitting your production.
**With Cyclops and our MCP, you can now abstract Kubernetes complexity from your developers AND your AI agents**
## Install via UI
The easiest way to install the Cyclops MCP server is through the Cyclops UI. Below are instructions on how to install it via binary and kubectl, but the recommended way of installing it is via the UI.
> ⚠️ To install Cyclops MCP through the UI, you should run Cyclops on a version `v0.20.1` or greater
1. Install the Cyclops MCP by going to sidebar `“Addon”` > `"MCP server"`. You can now click `"Install Cyclops MCP server"` and your Cyclops MCP server will be up and running in a few seconds.
<img width="762" alt="addon-install" src="https://github.com/user-attachments/assets/57dfd8f9-9b58-4797-a348-d2788a6fba9b" />
2. Now that your MCP server is up running, all thats left is exposing it outside of your cluster and connecting your AI agent to it.
To expose the MCP server on localhost you can simply port-forward it with the following command:
```bash
kubectl port-forward -n cyclops svc/cyclops-mcp 8000:8000
```
Your server is now available on `localhost:8000`.
3. To connect it to an AI agent you will just have to provide the Cyclops MCP server in its configuration. For example, to add it to Cursor, you can simply add it with the following JSON:
```bash
{
"mcpServers": {
"cyclops-kubernetes": {
"url": "http://localhost:8000/sse"
}
}
}
```
You can now start a fresh conversation with your AI companion. Above is an example of how we used it with Cursor.
<details>
<summary>Install stdin binary</summary>
### 1. Make sure Cyclops is installed in your Kubernetes cluster
Check our docs on how it install it with a single command - https://cyclops-ui.com/docs/installation/install/manifest
### 2. Download MCP server
You can download the Cyclops MCP server binary with the following command:
```yaml
GOBIN="$HOME/go/bin" go install github.com/cyclops-ui/mcp-cyclops/cmd/mcp-cyclops@latest
```
### 3. Add server configuration
> ⚠️ By default, Cyclops MCP will use the `.kube/config` file to connect to your cluster
Configure your MCP Cyclops server:
```json
{
"mcpServers": {
"mcp-cyclops": {
"command": "mcp-cyclops"
}
}
}
```
## Configuration
You can configure Cyclops MCP server via env variables. Below is an example of adding the configuration for specifying the kubeconfig file the Cyclops MCP server should use when managing your Cyclops applications.
```json
{
"mcpServers": {
"mcp-cyclops": {
"command": "mcp-cyclops",
"env": {
"KUBECONFIG": "/path/to/your/kubeconfig"
}
}
}
}
```
### Environment variables
Below is the list of environment variables used for configuring your Cyclops MCP server:
| Env var | Description |
|-----------------------------------|-----------------------------------------------------------------------------------------|
| `KUBECONFIG` | Path to kubeconfig file (optional, defaults to in-cluster config or $HOME/.kube/config) |
| `CYCLOPS_KUBE_CONTEXT` | Kubernetes context to use (optional) |
| `CYCLOPS_MODULE_NAMESPACE` | Namespace where modules are stored |
| `CYCLOPS_HELM_RELEASE_NAMESPACE` | Namespace for Helm releases |
| `CYCLOPS_MODULE_TARGET_NAMESPACE` | Target namespace for modules |
---
</details>
<details>
<summary>
Install to a Kubernetes cluster manually
</summary>
---
Instead of having each developer install `mcp-cyclops` binary, you can install the Cyclops MCP server with SSE as transport type to your Kubernetes cluster and allow all of your developers to connect to the same server.
1. Before installing, make sure Cyclops and all its CRDs are installed in your cluster:
1. Check Cyclops pods are running:
```shell
kubectl get pods -n cyclops
```
Should write:
```
NAME READY STATUS RESTARTS AGE
cyclops-ctrl-676b5d9789-ntcls 1/1 Running 0 94s
cyclops-ui-7798655f97-xdg29 1/1 Running 0 94s
```
2. Check if CRDs are installed
```shell
kubectl get crds | grep cyclops-ui
```
Should write:
```
modules.cyclops-ui.com 2025-04-26T15:28:18Z
templateauthrules.cyclops-ui.com 2025-04-26T15:28:18Z
templatestores.cyclops-ui.com 2025-04-26T15:28:18Z
```
2. Install Cyclops MCP server with the following command:
```shell
kubectl apply -f https://raw.githubusercontent.com/cyclops-ui/mcp-cyclops/refs/heads/main/install/mcp-server.yaml
```
3. You can now expose the `cyclops-mcp` service. To test your MCP server, you can port-forward it:
```shell
kubectl port-forward svc/cyclops-mcp -n cyclops 8000:8000
```
4. Add your Cyclops MCP server host, or in case you are testing it, the [localhost](http://localhost) address where you port-forwarded the MCP service:
```json
{
"mcpServers": {
"mcp-cyclops": {
"url": "http://localhost:8000/sse"
}
}
}
```
---
</details>
## Tools
| Tool | Description |
|-----------------------|------------------------------------------------------------------------------------------------------------------------------------|
| `create_module` | Create new Module. Before calling this tool, make sure to call `get_template_schema` to validate values for the given template |
| `get_module` | Fetch Module by name |
| `list_modules` | List all Cyclops Modules |
| `update_module` | Update Module by Name. Before calling this tool, make sure to call `get_template_schema` to validate values for the given template |
| `get_template_schema` | Returns JSON schema for the given template. Needs to be checked before calling `create_module` tool |
| `get_template_store` | Fetch Template Store by Name |
| `list_template_store` | List Template Stores from cluster |
