test_mcp_server

A simple MCP (Model Context Protocol) server project demonstrating both local and remote MCP server setups using FastMCP, LangChain, and uv.

Requirements

• Python 3.9+

• pip

• uv

Related MCP server: MCP Server Example

First-Time Setup

1. Install uv

pip install uv

2. Navigate to the project directory

cd test_mcp_server

3. Initialize the project with uv

uv init .

4. Add FastMCP

uv add fastmcp

Local MCP Server Setup

1. Create the local server file

Create a file named:

local_server.py

This file contains your MCP server implementation.

2. Add required dependencies

uv add langchain langchain-openai langchain_mcp_adapters

3. Create the client

Create a client file:

client.py

4. Run the local server using STDIO

The local MCP server communicates via STDIO.

Run the client:

uv run client.py

Remote MCP Server Setup

1. Create MCP tools

• Define your MCP tools for the remote server

• Ensure they are compatible with FastMCP Cloud

2. Deploy to FastMCP Cloud

• Deploy the server to FastMCP Cloud

• Obtain the remote server configuration

3. Update configuration

• Add the remote MCP server configuration to your config file

• Replace the local STDIO setup with the remote server endpoint

4. Run the client with the remote server

uv run client.py

Notes

• Local server uses STDIO for communication

• Remote server runs on FastMCP Cloud

• uv handles dependency management and execution

• Same client can be used for both local and remote servers by changing configuration

Remote MCP Server Deployment (FastMCP Cloud)

GitHub Access

• GitHub repository access was granted to FastMCP Cloud

• FastMCP Cloud pulls the source code directly from the repository

Deployment Steps

1. Created MCP tools for the remote server

2. Connected the GitHub repository to FastMCP Cloud

3. FastMCP Cloud executed the server using:

main.py

4. The server was successfully deployed as a remote MCP server

Client Configuration

• Updated the MCP configuration file to point to the remote FastMCP Cloud endpoint

• Reused the same client.py for both local and remote execution

Running the Client

uv run client.py