Glama
Sign In

Terminal MCP Server

by theailanguage
Verified
GitHub
4
  • Apple
RedditDiscord
OverviewInspectNewSchemaRelated ServersReviewsScore
Need Help?View Source CodeReport Issue

local-only server

The server can only run on the client’s local machine because it depends on local resources.

Integrations

  • Enables execution of shell commands in a dedicated workspace directory, returning command output or error messages to the AI

MCP Server for Executing Terminal Commands

Introduction

Hey everyone! Welcome back to The AI Language. In this project, we’re building an MCP server that can execute terminal commands. This allows you to ask Claude for Desktop to run commands on your computer and return the output — like using a terminal, but powered by AI!

🎥 Video tutorials:

  • Without Docker:
  • With Docker:

SUBSCRIBE

If you enjoy learning about AI, coding, and automation, please like and subscribe to the channel — it really helps us make more great content for you!

What is MCP?

MCP (Model Context Protocol) is a system that allows AI models to:

  • Store data (like files or API responses)
  • Run tools (functions that AI can execute)
  • Use prompts (predefined templates for tasks)

In this tutorial, we’re building a tool that takes a command, runs it in the terminal, and sends back the output.

Two Ways to Set Up Your MCP Server

You can run your MCP server with or without Docker:

🔹 Option 1: Run Directly Using Python (no Docker)

This is great for development or local experimentation.

🔹 Option 2: Run Using Docker

Perfect for clean environments and easy distribution — no need to install Python or dependencies globally.

Both methods work with Claude for Desktop, and we’ll walk you through both.

Option 1: Setup Without Docker (Local Python)

Step 1: Install Python and uv

You need Python 3.10 or higher and a tool called uv:

  • Install Python: https://www.python.org/downloads/
  • Install uv:
    curl -LsSf https://astral.sh/uv/install.sh | sh
  • Fix permissions if needed:
    sudo chown -R $(whoami):staff ~/.local

Step 2: Create Project Structure

mkdir -p ~/mcp/servers/terminal_server mkdir -p ~/mcp/workspace cd ~/mcp/servers/terminal_server

Step 3: Initialize and Activate Project

uv init uv venv source .venv/bin/activate # Mac/Linux # or .venv\Scripts\activate # Windows

Step 4: Install Required Packages

uv add "mcp[cli]"

Step 5: Create Your Server

Create terminal_server.py and paste this:

import os import subprocess import logging from mcp.server.fastmcp import FastMCP logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s") logging.info("Starting Terminal Server MCP...") mcp = FastMCP("terminal") DEFAULT_WORKSPACE = os.path.expanduser("~/mcp/workspace") logging.info(f"Workspace path: {DEFAULT_WORKSPACE}") @mcp.tool() async def run_command(command: str) -> str: logging.info(f"Running command: {command}") try: result = subprocess.run(command, shell=True, cwd=DEFAULT_WORKSPACE, capture_output=True, text=True) return result.stdout or result.stderr except Exception as e: logging.exception("Command failed") return str(e) if __name__ == "__main__": mcp.run(transport='stdio')

Step 6: Start the Server

uv run terminal_server.py

Option 2: Setup With Docker

Step 1: Create a Dockerfile

In the same folder as terminal_server.py, create a Dockerfile:

FROM python:3.11-slim WORKDIR /app COPY . /app RUN pip install --no-cache-dir -r requirements.txt EXPOSE 5000 CMD ["python", "terminal_server.py"]

Make sure you have a requirements.txt that includes:

mcp[cli]

Step 2: Build the Docker Image

docker build -t terminal_server_docker .

Step 3: Configure Claude Desktop to Use the Docker Image

Open:

code ~/Library/Application\ Support/Claude/claude_desktop_config.json

Add:

{ "mcpServers": { "terminal_server": { "command": "docker", "args": [ "run", "-i", "--rm", "--init", "-e", "DOCKER_CONTAINER=true", "-v", "/Users/<your-username>/mcp/workspace:/root/mcp/workspace", "terminal_server_docker" ] } } }

Replace <your-username> with your actual macOS username.

Then restart Claude Desktop.

Testing the MCP Server

In Claude, try these prompts:

  • Run the command ls in my workspace.
  • Execute echo Hello from Claude.

You should see the output directly from your terminal server 🎉

Wrapping Up

Congrats! You’ve built a working MCP server that executes terminal commands — and you can run it with or without Docker.

Want to go further?

  • Add security checks to block dangerous commands
  • Let Claude write to or read from files
  • Connect to cloud or remote systems

👉 Let us know in the YouTube comments what you'd like to build next!

Don’t forget to like, subscribe, and share to support more videos like this. See you next time 🚀

This server cannot be installed

-
security - not tested
F
license - not found
-
quality - not tested

An MCP server that allows AI assistants like Claude to execute terminal commands on the user's computer and return the output, functioning like a terminal through AI.

  1. Introduction
    1. What is MCP?
      1. Two Ways to Set Up Your MCP Server
        1. 🔹 Option 1: Run Directly Using Python (no Docker)
        2. 🔹 Option 2: Run Using Docker
      2. Option 1: Setup Without Docker (Local Python)
        1. Step 1: Install Python and uv
        2. Step 2: Create Project Structure
        3. Step 3: Initialize and Activate Project
        4. Step 4: Install Required Packages
        5. Step 5: Create Your Server
        6. Step 6: Start the Server
      3. Option 2: Setup With Docker
        1. Step 1: Create a Dockerfile
        2. Step 2: Build the Docker Image
        3. Step 3: Configure Claude Desktop to Use the Docker Image
      4. Testing the MCP Server
        1. Wrapping Up