cdp-tools

cdp-tools-mcp is an MCP server that brings Chrome DevTools Protocol-powered debugging and browser automation to AI agents. It enables runtime debugging features such as breakpoints, variable inspection, console and network monitoring, plus browser automation with element discovery and interaction recording. The server also supports multi-agent workflows, allowing nested agents to manage their own browser tabs within the same Chrome instance, and it includes server management capabilities like port monitoring and auto-restart. To use it, start the MCP via npx, then connect your MCP client (e.g., Claude or another MCP client) using the provided command or the config snippet. You can add the server to Claude Desktop configurations or run via stdio transport, after which you’ll be able to issue debugging commands, replay recorded interactions, and manage multiple sessions without port conflicts.

Prerequisites:

• Node.js and npm installed on your machine
• Internet access to fetch packages from npm

Installation steps:

1. Clone the repository (optional if you’re installing from npm): git clone https://github.com/InDate/cdp-tools-mcp.git cd cdp-tools-mcp
2. Install dependencies: npm install
3. Build the project (if applicable): npm run build
4. Start the MCP server: npm start
5. Optional: run directly via npx (as shown in the README): npx cdp-tools-mcp@latest

If you’re integrating with Claude, you can also add the server to your Claude config as shown in the Quick Start example.

Tips and notes:

• The MCP supports cross-session and cross-tool coordination. Enable auto-restart if you want servers to recover automatically after failures.
• Configure port monitoring and blocking behavior to control how the LLM reacts to server issues.
• Use the provided config snippet to integrate with Claude Desktop or other MCP clients; this keeps command and args centralized.
• For best results, ensure your environment allows Chrome DevTools Protocol connections and that any required Docker or container capabilities are available if you plan to run Docker-related workflows.
• The repository emphasizes recording and replaying reproduction steps for automated verification, as well as multi-agent tab management for parallel debugging sessions.