MCP Server Deployment for HTTP Serving - macOS Setup + Code

This project demonstrates how to build and run a local MCP server that exposes tools, and how to test those tools using MCP Inspector through a browser-based interface.

You'll define and run two simple tools on the server:

• One to search papers from arXiv based on a topic
• One to extract details about a specific paper

The project runs entirely on your local macOS system using Python 3.13.2.

The server is implemented using the Model Context Protocol (MCP), and the setup is already initialized with uv. After cloning the repository, you’ll create a virtual environment, install dependencies, and launch the MCP server and Inspector with a single command.

This is a self-contained example for learning how MCP servers are defined, how tools are registered, and how to interactively test them using the Inspector UI.

Project Workflow Overview

flowchart TD
    A[Clone Repository] --> B[Set Up Virtual Environment]
    B --> C[Install Project Dependencies]
    C --> D[Run MCP Server with Inspector]
    D --> E[Open Inspector in Browser]
    E --> F[Interact with MCP Tools Visually]

1. Clone and Set Up the Environment

Clone this repository:

git clone https://github.com/shaikhq/mcpworks.git
cd mcpworks

Create a virtual environment using Python 3.13:

uv venv --python /usr/local/bin/python3.13

Activate the environment:

source .venv/bin/activate

Confirm your Python version:

python --version
# Should show Python 3.13.x

2. Install Dependencies

uv add mcp arxiv

This will install the mcp runtime and arxiv client library required to run the server.

3. Run the MCP Server

uv run research_server_http.py

The above command will start the MCP server at the following endpoint: http://127.0.0.1:8000

!alt text

4. Run the MCP Inspector

From another terminal window.

Launch MCP Inspector without authentication. For most production deployments, you will likely need to setup authentication. However, in my current exercise, my focus was to deploy a MCP server with HTTP end point. So, I haven't enabled authentication yet.

export DANGEROUSLY_OMIT_AUTH=true
npx @modelcontextprotocol/inspector

!alt text

• MCP Inspector runs a local UI for browsing and testing tools
• The MCP server runs your tool code and handles requests

Once started, open your browser to:

http://127.0.0.1:6274/

You can now interact with your tools from the browser interface.

In the URL field of the inspector, type the HTTP URL of the MCP server and click connect. Then go to the tools tab to see the tools from the MCP server and try them out.

!alt text