> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mogenius.com/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Server

> Connect external AI tools and IDEs to your mogenius cluster using the Model Context Protocol (MCP) server.

mogenius exposes a Model Context Protocol (MCP) server that lets external AI-powered tools — such as IDE assistants, Claude Desktop, or custom automations — interact with your Kubernetes cluster through the same AI infrastructure used by [AI Chat](/ai/chat) and [AI Insights](/ai/ai-insights).

## Setup

The MCP server is available through the mogenius API and requires a mogenius API key for authentication.

## Authentication & Permissions

Access to the MCP server requires a valid **API key** generated through the mogenius platform. Each API key is tied to a specific user and inherits that user's workspace permissions.

This means:

* The MCP client can only access namespaces and resources the API key owner has permission to view or manage.
* For different Workspaces you can create individual API keys, each scoped to the resources and permissions in a Workspace.
* API keys can be revoked at any time from the platform settings.

You can create an API key from the settings of a **Cluster** or **Workspace**.

### Connecting a Client

Add the mogenius MCP server to your client's configuration. The exact steps depend on the tool you're using, but typically you can edit a local config file that holds the MCP servers of your client.

#### Claude Desktop

For remote MCP servers Claude typically refers to Connectors. mogenius is not yet available as a Connector in Claude but you can configure it as a local MCP server. Modify the following JSON to match your setup and add it to your `claude_desktop_config.json` file.

```
"mcpServers": {
    "mogenius": {
      "command": "PATH-TO-YOUR-NODE-INSTALLATION/.nvm/versions/node/v23.3.0/bin/npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://platform-api.mogenius.com/mcp",
        "--header",
        "Authorization: bearer YOUR-MOGENIUS-API-KEY"
      ],
      "env": {
        "PATH": "PATH-TO-YOUR-NODE-INSTALLATION/.nvm/versions/node/v23.3.0/bin:/usr/local/bin:/opt/homebrew/bin:/usr/bin:/bin"
      }
    }
  },
```

#### Cursor

In your Cursor Desktop app open **Settings** and navigate to **Tools & MCP**. Add a new MCP server. Cursor will open your mcp.json and you can add the mogenius MCP server like this.

```
{
  "mcpServers": {
    "mogenius": {
        "type": "streamable-http",
        "url": "https://platform-api.mogenius.com/mcp",
        "headers": {
          "Authorization" : "bearer YOUR-MOGENIUS-API-KEY"
      }
    }
  }
}
```

## Tools & Capabilities

The MCP server exposes the same tools available in AI Chat, allowing external clients to perform cluster operations through natural language or structured requests.

### Kubernetes

External clients can query and inspect Kubernetes resources through the MCP server:

* List and describe Pods, Deployments, Services, ConfigMaps, and other resources.
* Read Pod logs and events for troubleshooting.
* Retrieve resource manifests and status information.
* Inspect namespace-level overviews.
* Create and modify resources and workloads.

### Helm

The MCP server provides access to Helm operations:

* List installed Helm releases across namespaces.
* Inspect chart values and release configuration.
* Review release history and rollback information.
* Deploy Helm charts.

### GitHub

When GitHub is connected in [AI Setup](/ai/setup#github), the MCP server can also access linked repositories:

* Browse repository contents, files, and directory structures.
* Read source code, configuration files, and CI/CD pipelines.
* Cross-reference cluster resources with the codebase that manages them.

If a [Memory Repository](/ai/setup#memory-repository) is configured, the MCP server can read and write persistent context, enabling external tools to benefit from the same long-term knowledge base used by AI Chat and AI Insights.