Crowpus MCP Client Bridge & CLI Setup Tool

A lightweight, universal CLI tool and stdio-to-SSE bridge for the Crowpus Model Context Protocol (MCP) Server.

This tool allows local AI agents (such as Claude Desktop, Cursor, Cline, Roo Code, and Antigravity) to easily connect to the secure, remote Crowpus MCP Server. It securely stores credentials locally and starts a standard input/output (stdio) bridge to interact with the remote server on behalf of the agent, automatically injecting authentication headers.

---

Features

• Secure Login: Saves your personal crow_mcp_... token locally in ~/.config/crowpus/config.json with secure 0600 permissions.
• Stdio-to-SSE Bridge: Bridges the remote EventSource/SSE transport of mcp.crowpus.dev to a local process using stdio transport (standard JSON-RPC over stdin/stdout).
• Auto-Integration Setup: Scans your local directories and automatically configures MCP settings for:
  • Claude Desktop
  • Cursor
  • Cline (VS Code extension)
  • Roo Code (VS Code extension)

---

Installation

Prerequisites

• Node.js (v18 or higher)
• npm

Setup & Build

Clone the repository and install the dependencies:

# Clone the repository
git clone https://github.com/CrowPus/mcp-cli.git
cd mcp-cli

# Install dependencies
npm install

# Compile TypeScript to JavaScript
npm run build

---

Usage

1. Log In (Configure Credentials)

Save your Crowpus MCP token securely to your machine:

node dist/index.js login <crow_mcp_token>

2. Auto-configure Platforms

Automatically scan and setup all detected AI agent environments on your machine:

node dist/index.js setup

This updates settings files (e.g. claude_desktop_config.json, cline_mcp_settings.json, and global-settings.json) to register the local command bridge:

{
  "mcpServers": {
    "crowpus": {
      "command": "npx",
      "args": ["-y", "@crowpus/mcp-cli", "run"]
    }
  }
}

3. Run Stdio Bridge Manually

If you want to run the stdio bridge process directly:

node dist/index.js run

---

How It Works

AI agent platforms natively support connecting to MCP servers that run as local child processes over standard input/output (stdio). However, they usually do not support connecting to remote SSE servers that require dynamic custom headers (like Authorization: Bearer <token>).

The @crowpus/mcp-cli acts as a local proxy.

1. The AI Agent boots the bridge command (npx @crowpus/mcp-cli run).
2. The local bridge reads the stored token and opens a secure connection to https://mcp.crowpus.dev/mcp using Streamable HTTP/SSE.
3. It fetches the remote tools and registers them locally.
4. When the agent calls a tool, the bridge forwards the call to the remote cloud server and returns the results.