claude-code-bridge

This MCP server acts as a bridge between Claude Desktop (stdio) and the Chrome Extension (HTTP/SSE) to enable a unified, agentic workflow. It exposes seven tools for inter-agent coordination: task delegation to Claude Code CLI, an agent message board for async communication, a command executor with built-in safety checks, file reading with path protections, a system status monitor, and a skills inventory explorer. The server supports two transport channels: stdio for Claude Desktop and HTTP/SSE for the Chrome Extension, allowing you to run Claude Desktop locally while interacting with the browser extension for streamable messaging and health checks. You can query status, send tasks to Claude Code, read messages from teammates, and discover installed skills all through the same MCP interface.

To use it, ensure Claude Code CLI is installed and available in your environment, and connect Claude Desktop to the stdio bridge or the Chrome Extension to the HTTP bridge. The server exposes endpoints such as /health, /mcp for streamable transport, and /sse for legacy SSE via the Chrome Extension, enabling both real-time and asynchronous interactions across agents. With the Ralph autonomous loop and hooks in place, Claude Code can run tasks in a semi-autonomous development cycle, using the seven tools to coordinate work, share context, and manage execution across environments.

Prerequisites

• Node.js 18+ installed
• Claude Code CLI installed (npm install -g @anthropic-ai/claude-code)
• Claude Desktop installed (for stdio bridge) and/or Chrome Extension installed (for HTTP bridge)

Installation steps

1. Clone the repository and install bridge dependencies:

git clone https://github.com/YOUR_USER/claude-code-bridge.git
cd claude-code-bridge/bridge
npm install

2. Start the stdio bridge (Claude Desktop):

# Depending on your environment, point to your local path
node server.js

3. Start the HTTP bridge (Chrome Extension):

# Start the HTTP server that clients (Chrome Extension) connect to
bash start-http.sh

4. Configure Claude Desktop to connect to the stdio bridge (example):

• Windows: Edit %APPDATA%/Claude/claude_desktop_config.json to point to the stdio bridge { "mcpServers": { "claude-code-bridge": { "command": "wsl.exe", "args": ["-e", "bash", "/path/to/claude-code-bridge/bridge/start.sh"] } } }
• Mac/Linux (no WSL): { "mcpServers": { "claude-code-bridge": { "command": "node", "args": ["/path/to/claude-code-bridge/bridge/server.js"] } } } Restart Claude Desktop after configuring.

Tips and common issues:

• If you see blocked commands or path sanitization messages, verify the security settings in the server and ensure only allowed commands/paths are executed.
• The Tools Reference exposes seven tools; you can call them via the MCP interface or through Claude Desktop/Chrome Extension actions. Use system_status to monitor memory, Docker, disk space, and services.
• For the HTTP bridge, use http://localhost:3899/mcp for streamable transport and http://localhost:3899/sse for legacy SSE if needed. Ensure your firewall allows port 3899.
• Environment variables can be extended in the mcp_config under env to control runtime behavior, such as logging levels or feature flags.
• If you run into connection issues between Claude Desktop and the bridge, verify that the correct transport (stdio vs HTTP) is configured and that the start scripts have executable permissions (chmod +x start.sh start-http.sh).
• The npm_package field is set to claude-code-bridge to reflect its Node.js server nature, but this will vary if you deploy via a different packaging method.