Name: Terminal MCP Server
Author: weidwonder

LABSCO SUMMARY

The smallest surface there is: one tool, a command, and a directory to run it in.

There is nothing to learn — the interface is a command string and where to run it — which makes this the general-purpose fallback for the many jobs that have no dedicated server. The tool describes security restrictions without enumerating which, so treat the boundary as yours to establish before pointing an agent at a machine that matters.

What it is

A one-tool server that executes a terminal command in a specified working directory, with security restrictions applied by the server.

What you get

Any shell command executed on the machine running the server, through a single call
A workingDir argument on that call, so the command runs where you meant it to rather than wherever the server started

Requirements

No account and no key. What it does need is a machine you are content to let it run commands on: the tool takes a command and a workingDir and executes them locally.

Setup effort

Build from source — git clone https://github.com/weidwonder/terminal-mcp-server.git

See installation options

USE CASES

01

Running the build or the tests without leaving the conversation

Give the command and the directory it belongs in, and read the output where you were already working rather than switching to a shell and back.

02

Checking the state of a machine before you change anything on it

Run the read-only commands you would run by hand — what is installed, what is running, what is in this directory — with workingDir pinning each one to the right place.

1 tool

Listed in project docs. Your client may show a different list.

execute_command

BEFORE YOU INSTALL

What to know first

Tools you get

1 tool named in the project's documentation. execute_command has no allowlist: it runs any command as the user who started the server, or on a remote machine over SSH. A session keeps terminal state for 20 minutes.

INSTALLATION

Connect your client

You clone and build this one yourself.

Server setup

TERMINAL

git clone https://github.com/weidwonder/terminal-mcp-server.git
cd terminal-mcp-server
npm install
npm run build

Add the stdio mcpServers block to your client's settings file (Roo Code: ~/Library/Application Support/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/cline_mcp_settings.json; Cline: ~/.cline/config.json; Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json), pointing args at your built build/index.js.

For remote command execution, SSH key-based authentication must be set up in advance

MCP Clients

Claude Desktop, Roo Code, Cline

JSON

{
  "mcpServers": {
    "terminal-mcp": {
      "command": "node",
      "args": [
        "<YOUR_PATH_TO_TERMINAL_MCP_SERVER>/build/index.js"
      ],
      "env": {}
    }
  }
}

Roo Code — SSE, Cline — SSE

Start the server in SSE mode from your built clone:

TERMINAL

node build/index.js --sse

Then add this block. 8080 and /sse are the defaults; change them if you started the server with --port or --endpoint: