start

Run the CSP Analyser MCP server over stdio. This is how AI coding agents (Claude Code, Cursor, etc.) invoke CSP Analyser as a tool.

Usage

csp-analyser start

The command takes no options or arguments. Once started, the process speaks the Model Context Protocol over stdin/stdout and stays running until the transport closes.

You rarely run this directly — agents launch it for you via their MCP server configuration. See the MCP Server guide for full details on which tools are exposed, and MCP Configuration for how to register the server in Claude Code, Cursor, and other clients.

Example MCP configuration

Claude Code's ~/.claude.json entry for the published CLI:

{
  "mcpServers": {
    "csp-analyser": {
      "command": "csp-analyser",
      "args": ["start"]
    }
  }
}

Or via the Claude Code CLI:

claude mcp add csp-analyser -s user -- csp-analyser start

Pointing at a local build

For development — run the server directly from the compiled output without installing globally:

claude mcp add csp-analyser -s user -- node /path/to/CSPAnalyser/dist/mcp-server.js

The start subcommand is a thin wrapper that calls dist/mcp-server.js, so both forms are equivalent when CSP Analyser is installed.

Related

• MCP Server Overview — what the server does and when to use it
• MCP Tools Reference — the full list of tools exposed over the MCP transport
• MCP Configuration — client-specific setup

When to use this command

Use start when you want AI coding agents to use CSP Analyser as a tool. Running csp-analyser start launches the MCP server over stdio, allowing agents like Claude Code and Cursor to invoke CSP analysis tools programmatically. You typically don't run this command directly — instead, configure it in your agent's MCP settings. See the MCP Configuration guide for setup instructions.

Related commands

• crawl — The CLI equivalent of the MCP crawl_url tool
• setup — Install the browser before starting the MCP server