MCP Server Commands
Let LLMs run shell commands and processes on the host machine via a single run_process tool.
Add to your client
Copy the config for your MCP client and paste it into its config file.
npx mcp-server-commandsPaste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"mcp-server-commands": {
"command": "npx",
"args": [
"mcp-server-commands"
]
}
}
}Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- Node.js (npx) to run the published mcp-server-commands npm package
- An MCP client such as Claude Desktop, Zed, or Groq Desktop
About MCP Server Commands
An MCP server providing a single run_process tool that lets an LLM execute shell commands or invoke executables directly on the host machine, returning STDOUT and STDERR. Supports optional stdin for piping scripts or creating files, and ships a run_process prompt template. Communicates over STDIO.
Tools & capabilities (1)
run_processRuns a process on the host machine. Invoke via command_line (string, executed through the system default shell — supports pipes, redirects, variable expansion) or argv (string array, direct executable invocation with no shell interpretation); the two are mutually exclusive. Returns STDOUT and STDERR as text. Accepts an optional stdin parameter to pipe scripts into commands like fish/bash/zsh/python or to create files via cat.
When to use it
- Run quick host commands like hostname, ls -al, or echo from within an LLM chat
- Pipe scripts over stdin to interpreters such as bash, zsh, fish, or python
- Create files by feeding text through stdin to commands like cat >> path
- Drive local automation and dev workflows agentically with a tool-capable model
Security notes
Runs arbitrary commands with the permissions of the user that starts the server. Be careful what you ask it to run. In Claude Desktop use 'Approve Once' (not 'Allow for This Chat') so you can review each command, and 'Deny' for untrusted commands. DO NOT run the server with sudo. If bridging to HTTP via mcpo, vet it for security concerns.
MCP Server Commands FAQ
Can I pass both command_line and argv?
No. They are mutually exclusive — the tool infers whether to use a shell based on which parameter you provide. command_line goes through the system's default shell; argv invokes the executable directly with no shell interpretation.
How do I expose this over HTTP instead of STDIO?
The server is implemented with the STDIO transport. For HTTP/OpenAPI, bridge it with mcpo (e.g. 'uvx mcpo --port 3010 --api-key "supersecret" -- npx mcp-server-commands'), which works with Open-WebUI. Vet mcpo for security concerns before relying on it.
Is it safe to run?
It executes arbitrary commands with the permissions of the user running the server, so review each command. In Claude Desktop use 'Approve Once' to review every command and 'Deny' for untrusted ones, and never run the server with sudo.
Alternatives to MCP Server Commands
Compare all alternatives →
Microsoft's official browser-automation MCP using Playwright's accessibility tree (no vision model).
Up-to-date, version-specific library documentation injected into your coding agent.
LSP-powered coding agent toolkit: semantic symbol search, references and structural edits.
Compare MCP Server Commands with: