Kubescape Mcp

What is Kubescape Mcp?

Kubescape Mcp is an MCP (Model Context Protocol) server that exposes kubescape CLI commands as tools for AI assistants. It is designed for users of Claude Desktop or Cursor who want to perform Kubernetes security scans—such as image vulnerability scanning—directly through natural language prompts. The server requires the kubescape CLI and uv to be installed beforehand.

How to use Kubescape Mcp?

First, install the kubescape CLI and uv. Then add the following JSON configuration to your MCP settings in Claude Desktop or Cursor, adjusting the directory path to point to your local copy of the server:

{
  "mcpServers": {
    "kubescapeserver": {
      "command": "uv",
      "args": ["--directory", "/Users/haji/mcp-servers/kubescapeserver", "run", "server.py"]
    }
  }
}

After configuration, the server’s tools become available and can be invoked via prompts (e.g., “scan image nginx”).

Key features of Kubescape Mcp

• Exposes 4 tools that map to kubescape CLI commands.
• Scans container images for vulnerabilities.
• Provides AI-generated security suggestions.
• Supports security frameworks like NSA and MITRE ATT&CK.
• Works with both local and public container images.

Use cases of Kubescape Mcp

• Scanning a custom container image for vulnerabilities during development.
• Auditing a public image (e.g., nginx) for known security issues.
• Obtaining framework-specific compliance advice (NSA, MITRE ATT&CK) after a scan.
• Automating Kubernetes security checks through an AI assistant interface.

FAQ from Kubescape Mcp

What are the prerequisites for using Kubescape Mcp?

You must have the kubescape CLI installed (see https://kubescape.io/docs/getting-started/) and the uv package manager installed on your machine.

How do I configure Kubescape Mcp for my client?

Add the JSON configuration block shown in the README to the MCP settings of Claude Desktop or Cursor, replacing the directory path with the actual location of the server script.

What tools does Kubescape Mcp expose?

It exposes 4 tools that are direct wrappers around kubescape CLI commands. The README does not list their names explicitly.

Can I scan a local image that is not present in my cluster?

The README shows that if an image test is not found, the model suggests switching to a public image scan (e.g., nginx) instead.

What security frameworks are supported by Kubescape Mcp?

After an image scan, AI suggestions are provided for specific frameworks including NSA and MITRE ATT&CK. The README mentions “other specific frameworks” but does not list an exhaustive set.