mcp-server-commands

Summary

This repository provides tools for Large Language Models (LLMs), specifically designed for use with the Claude Desktop application. It introduces two primary tools: run_command, which executes shell commands and returns STDOUT and STDERR, and run_script, which executes scripts via STDIN, enabling LLMs to run code they generate.

The README emphasizes caution when using these tools due to potential security risks, recommending the "Approve Once" setting in Claude Desktop. It includes a video walkthrough for better understanding.

The repository also provides prompts for users to include in chat history.

For development, it outlines steps for installing dependencies (npm install), building the server (npm run build), and using auto-rebuild during development (npm run watch).

Installation instructions detail how to configure the server in Claude Desktop using either the published npm package (mcp-server-commands) or a local build. It also covers logging, explaining how to enable verbose logging with the --verbose flag. For debugging, it recommends using the MCP Inspector, accessible via npm run inspector.