MCP_Server_Demo by stefano-chen - MCP Server

MCP Server Demo

A demo implementation of a Model Context Protocol (MCP) server using FastAPI and the mcp-sdk for Python. This server is designed as a lightweight reference implementation to support advanced agentic RAG systems and other MCP-based architectures.

This custom MCP server is designed to work with this project

About
Features
Architecture & Components
Getting Started
- Prerequisites
- Installation

About

MCP Server Demo is a reference server for the Model Context Protocol (MCP), built in Python using the mcp-sdk.
It's intended for experimentation, learning, and integration with more complex systems (e.g. agentic RAG setups).

Features

HTTP interface (via FastAPI) to accept MCP requests
STDIO mode for simpler or local testing
Minimal dependencies and clear code paths

Architecture & Components

Component	Description
`api_server.py`	Primary HTTP interface (FastAPI) handling service endpoints
`mcp_server_http.py`	Custom MCP server using the HTTP transport mode
`mcp_server_stdio.py`	Custom MCP server using the STDIO transport mode
`requirements.txt`	Dependencies and version constraints

Getting Started

Prerequisites

Python 3.9+
pip
(Optional) Virtual environment tool, e.g. venv or virtualenv

Installation

Clone the repo:

    git clone https://github.com/stefano-chen/Advanced_Agentic_RAG.git
    cd Advanced_Agentic_RAG

Create and activate a venv:

python3 -m venv .venv
source .venv/bin/activate      # macOS / Linux
.\.venv\Scripts\activate       # Windows (PowerShell / cmd)

Install requirements:
```
pip install -r requirements.txt
```
Launch the FastAPI server
```
fastapi run api_server.py
```
This will launch the MCP server on the default port 8000. The MCP server will interact with this service using the endpoints
Running the MCP server in HTTP mode
```
python mcp_server_http.py
```
This will allow external LLM to access the MCP server via HTTP protocol

stefano-chen/MCP_Server_Demo