Secure MCP servers with cnspec

Scan a Model Context Protocol (MCP) server to understand the capabilities it exposes to AI agents and LLM clients. cnspec connects to an MCP server and inventories its tools, prompts, resources, and resource templates so you can review what an agent connected to the server is able to do.

Prerequisites

To scan an MCP server with cnspec, you must have:

• cnspec installed on your workstation
• An MCP server you are authorized to scan, reachable over HTTP or runnable as a local command

Connect to an MCP server

cnspec connects to an MCP server using one of two transports:

• HTTP for remote servers that expose a streamable HTTP endpoint
• stdio for local servers that run as a command

To test access over HTTP, open a cnspec shell:

cnspec shell mcp http https://mcp.example.com/mcp

To connect to a local server over stdio, pass the command to run as a single quoted argument:

cnspec shell mcp stdio "npx -y @modelcontextprotocol/server-everything"

Verify the connection by listing the tools the server exposes:

cnspec> mcp.tools { name }
mcp.tools: [
  0: {
    name: "search_documents"
  }
  1: {
    name: "create_issue"
  }
  ...
]

Scan an MCP server

To scan an MCP server over HTTP:

cnspec scan mcp http https://mcp.example.com/mcp

To scan a local MCP server over stdio:

cnspec scan mcp stdio "npx -y @modelcontextprotocol/server-everything"

Understand scan output

When a scan completes, cnspec prints a summary of all the checks it ran, grouped by policy. Each check shows a pass or fail result. For example:

✓ Pass:  Ensure every exposed tool documents its behavior
✕ Fail:  Ensure tool input schemas declare required parameters
✓ Pass:  Ensure prompts include a description

At the end of the output, cnspec shows a risk score from 0 (no risk) to 100 (highest risk). Failed checks include remediation guidance to help you fix issues.

Scan with the Mondoo MCP Security policy

Mondoo maintains an out of the box Mondoo Model Context Protocol (MCP) Security policy that reviews the tools, prompts, and resources an MCP server exposes.

Mondoo Platform users: Enable the policy in your space. In the Mondoo Console, go to Findings > Policies, search for "MCP", and add the policy. All future scans of your MCP servers automatically evaluate against it. To learn more, read Manage Policies.

Open source users: Pass the policy bundle URL directly to cnspec:

cnspec scan mcp http https://mcp.example.com/mcp \
  --policy-bundle https://raw.githubusercontent.com/mondoohq/cnspec/refs/heads/main/content/mondoo-mcp-security.mql.yaml

You can also create your own policies to meet your specific requirements.

Explore an MCP server

Run cnspec shell mcp http https://mcp.example.com/mcp to open the interactive shell.

Inspect exposed tools

cnspec> mcp.tools { name description }
mcp.tools: [
  0: {
    name: "search_documents"
    description: "Search the knowledge base for matching documents"
  }
  ...
]

Review tool input schemas

cnspec> mcp.tools { name inputSchema { type required } }

List prompts and resources

cnspec> mcp.prompts { name title description }
cnspec> mcp.resources { name title description }

Example security checks

Ensure every tool documents its behavior

cnspec> mcp.tools.all( description != "" )

Ensure every prompt includes a description

cnspec> mcp.prompts.all( description != "" )

Find tools whose input schema declares no required parameters

cnspec> mcp.tools.where( inputSchema.required.length == 0 ) { name }

Learn more

• To learn more about how the MQL query language works, read Write Effective MQL.

• To learn about all the MCP resources and properties you can query, read the Mondoo MCP Resource Pack Reference.