Overview

Let AI assistants like Claude and Cursor work with your Intruder account in plain English.

Intruder MCP Server

Let AI assistants like Claude and Cursor work with your Intruder account in plain English. Add targets, start and schedule scans, triage issues, check your attack surface and license usage, and write management summaries, all without leaving your AI assistant.

The Intruder MCP server is open source. It connects your AI assistant to the Intruder API using the Model Context Protocol (MCP), an open standard for connecting AI assistants to the tools you already use.


How it works

The MCP server gives your AI assistant a set of tools, each one an Intruder action such as listing issues or starting a scan. When you ask a question, the assistant reads each tool's description and picks the ones it needs.

The server runs on your own computer and only talks to the Intruder API, using your API access token. Some tools make changes to your account, such as adding or deleting targets, starting scans or snoozing findings. Most AI assistants ask you to approve each tool call before it runs, so you stay in control.


Before you start

You'll need:

  • An Intruder account on a Cloud, Pro, Enterprise or Vanguard plan
  • An Intruder API access token, which only account admins can create: see Creating an access token
  • An AI assistant that supports local MCP servers, such as Claude Desktop, Claude Code or Cursor. Claude in a web browser (claude.ai) isn't supported.

Treat your access token like a password: anyone with it can see your targets and issues, and make changes to your account.


Set up Claude Desktop (recommended)

The easiest way to get started. There's nothing else to install.

  1. Download intruder.mcpb.
  2. Double-click the file (or drag it into Settings → Extensions in Claude Desktop) and click Install.
  3. Paste your API access token when asked.

Claude Desktop stores your token encrypted and runs the server for you.

Note: Claude Desktop warns that the extension can access everything on your computer and that its developer details aren't verified by Anthropic. It shows this for every extension installed from a file. The Intruder extension only talks to the Intruder API (api.intruder.io), and its source code is public.


Set up Claude Code, Cursor and other clients

These clients run the server with uv, a small tool that downloads Python and the server for you. Install uv first, then follow the steps for your client. Replace your-api-key with your API access token.

Claude Code

Run this in your terminal:

claude mcp add intruder -e INTRUDER_API_KEY=your-api-key -- \
  uvx --from git+https://github.com/intruder-io/intruder-mcp intruder-mcp

Cursor and other MCP clients

Add this to your client's MCP configuration (for Cursor, ~/.cursor/mcp.json):

{
  "mcpServers": {
    "intruder": {
      "command": "uvx",
      "args": ["--from", "git+https://github.com/intruder-io/intruder-mcp", "intruder-mcp"],
      "env": {
        "INTRUDER_API_KEY": "your-api-key"
      }
    }
  }
}

Some apps don't see the same PATH as your terminal. If the server doesn't start, replace uvx with its full path, which you can find by running which uvx (macOS and Linux) or where uvx (Windows).

Prefer Docker? The server is also published as a container image, ghcr.io/intruder-io/intruder-mcp. See the README for Docker setup, manual Claude Desktop configuration and troubleshooting.


Check it works

Ask your AI assistant:

"Who am I authenticated as in Intruder?"

It should reply with the email address of the account your access token belongs to. Then try some use cases.


Intruder MCP server or Developer Hub MCP server?

This Developer Hub also has its own MCP server, at https://developers.intruder.io/mcp. It lets coding assistants look up Intruder API endpoints and schemas while you write code against the API. It doesn't connect to your Intruder account.

To have an AI assistant work with your Intruder account, use the Intruder MCP server described on this page.


Further resources


Did this page help you?