# Interact with Your Rafay Managed Kubernetes Clusters Using MCP-compatible AI clients

February 25, 2026

.png)

\
\
Ankur Pandita

The **Model Context Protocol (MCP)** is an open standard that enables AI assistants to securely interact with external tools and systems. When used with Kubernetes, MCP allows an AI assistant to execute operations (for example, `kubectl` commands), retrieve live cluster state, and reason about results without requiring users to manually copy and paste output into a chat interface.

This blog uses Claude Desktop as an example AI assistant. The same approach applies to any MCP-compatible AI client.

For platform administrators, this capability enables controlled, auditable, and policy-driven AI-assisted cluster operations.

### Recommended Architecture: Local MCP Server with Rafay ZTKA Kubeconfig

For production environments, the recommended approach is to run the MCP server locally and connect to your Kubernetes cluster using a **Rafay Zero Trust Kubectl Access (ZTKA) kubeconfig**.

In this model:

- The MCP server runs on the administrator’s workstation
- Cluster access is established through Rafay’s ZTKA secure relay
- No inbound access to the cluster is required
- No VPN tunnels or exposed Kubernetes API endpoints are needed

This architecture aligns with zero-trust security principles and enterprise compliance requirements.

#### Security and Governance Considerations for Platform Admins

When integrating AI-driven access into Kubernetes environments, security, identity, and auditability must remain fully enforced. Rafay ZTKA ensures:

.png)

#### 1. Authentication (AuthN) and Authorization (AuthZ)
- Access is tied to verified user identity
- Authorization is enforced via Rafay RBAC policies
- No static tokens or long-lived credentials are required
- Permissions are evaluated for every request

The AI assistant does not bypass cluster security controls, it operates strictly within the RBAC boundaries of the authenticated user.

#### 2. Audit Logging
- Every `kubectl` request routed through ZTKA is recorded in the Rafay platform
- All actions initiated via the MCP server are fully auditable
- Logs can be used for compliance validation, forensics, and operational review

This ensures AI-assisted operations are as traceable as manual administrative actions.

#### 3. RBAC-Controlled Access
- Access to clusters, namespaces, and resources is governed by Rafay RBAC
- Platform teams can restrict AI-assisted access to specific roles or environments
- Fine-grained access control remains intact

#### 4. No Exposed Cluster Endpoints
- ZTKA uses a secure relay architecture
- Kubernetes API servers do not need to be publicly accessible
- No direct inbound network exposure is introduced

### Prerequisites

Before enabling MCP-based Kubernetes access, ensure the following components are installed and configured:

- **`mcp-server-kubernetes` (installed globally):**
  
  `npm install -g mcp-server-kubernetes`

- A **ZTKA kubeconfig file** downloaded from the Rafay Console
- **`kubectl`** installed locally
- An **MCP-compatible AI client** (Claude Desktop is used here as an example)

Installing `mcp-server-kubernetes` globally ensures the executable is available in your system `PATH`, allowing your AI client to invoke it correctly.
- Configure the AI assistant (e.g. Claude Desktop)

```javascript
{
  "mcpServers": {
    "kubernetes": {
      "command": "mcp-server-kubernetes",
      "env": {
        "KUBECONFIG": "/path/to/ztka-cluster-config.yaml"
      }
    }
  }
}
```

Replace `/path/to/ztka-cluster-config.yaml` with the actual path to your ZTKA kubeconfig.

### Connecting Your AI Client (Example: Claude Desktop)

After configuring the MCP server to use your ZTKA kubeconfig:

- Restart your AI client
- Confirm that Kubernetes tools appear in the client’s connectors or tool menu

Start a new session and select the Kubernetes integration if prompted

Once connected, the AI assistant can securely execute Kubernetes commands through the MCP server.

### Validate the Integration

To verify the setup, try simple test commands such as:
- `List all pods in all namespaces`
- `Fix the pods or resources which are in error state of crashloop back state`

On first use, your AI client will request permission to execute Kubernetes operations. Approve the request to continue.

### Operational Recommendations for Platform Teams

Before rolling out this capability broadly:
- Review and validate RBAC permissions
- Restrict write access where not required
- Pilot the integration in non-production environments
- Monitor audit logs during the initial rollout
- Establish governance guidelines for AI-assisted operational workflows

### Summary

By combining MCP with Rafay ZTKA, organizations can enable AI-driven Kubernetes interactions without compromising security, visibility, or compliance.

This integration provides:
- Identity-based access control
- RBAC enforcement
- Full auditability
- A zero-trust network posture

While this guide demonstrates the workflow using Claude as an example AI client, the same architecture applies to any MCP-compatible assistant.

### What's Next

We are developing a native **Rafay MCP Server** that will expose Rafay-specific discovery and action-oriented capabilities through MCP including multi-cluster operations, add-on and blueprint management, and more. Stay tuned for updates.