MCPTutorial

Prerequisite

Please clone the repo on local

1. Install python >= 3.10

Select add python to path

Then restart vscode

2. Install python >= 3.10

run in vscode terminal

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

restart vscode

3. Run python environment

Run below commands inside the MCP folder

# Create virtual environment and activate it
uv venv
.venv\Scripts\activate

# if you encounter error like activate.ps1 cannot be      
loaded because running scripts is disabled on this system. For more information, see about_Execution_Policies, run below command
Set-ExecutionPolicy -Scope CurrentUser -ExecutionPolicy RemoteSigned

# Install dependencies
uv add mcp[cli] httpx Pillow fastmcp openai azure.identity

# You may need to update to latest MCP version
uv pip install --upgrade fastmcp

# You can run the server stand alone in terminal by below command
uv run python server.py

4. Set up MCP server

Open user setting by ctrl + shift + P

Paste below to replace Change the filepath to the local absolute path for the main.py type open user setting json and paste below For local, change the server.py code in main

mcp.run()

then change the open user setting to below

{
    "mcp": {
        "servers": {
            "mcpTest": {
                "type": "stdio",
                "command": "uv",
                "args": [
                    "run",
                    "--with",
                    "mcp[cli]",
                    "--with",
                    "pillow",
                    "--with",
                    "fastmcp",
                    "mcp",
                    "run",
                    "C:\\Users\\lipan\\Documents\\GitHub\\MCPTutorial\\MCP\\server.py"
                ]
            }
        }
    }
}

For SSE change server code in server.py on main to

mcp.run(transport="sse", host="127.0.0.1", port=9000)

and change open user setting to below

{
    "mcp": {
        "servers": {
            "mcpTest": {
                "type": "sse",
                "url": "http://localhost:9000/sse"
            }
        }
    }
}

4. Run MCP server

click start and it will start running

sign in github copilot and select claude 3.7

click tool in vscode agent and select tools

You can ask something like

whats the weather in kirkland today

To debug mcp server Run below command

uv run mcp dev main.py
# if you see the error, npx not found. Please ensure Node.js and npm are properly installed and added to your system PATH.
# then you need to install the latest nodejs https://nodejs.org/en/download

5. Run MCP client

To run the MCP Client Please make sure you are added in the security group(is adding all people currently) Then type below command to log into azure first

az account set --subscription "fe105136-f441-4214-8a12-1d1f4955e15f"
az login 
then press enter to continue

Need to change below for different protocol testing in client.py

# Create MCP client

# Local STDIO
client = Client(
    "server.py", 
    sampling_handler=sampling_handler, 
    log_handler=log_handler,
    roots=[
        "file://home/projects/roots-example/frontend"
    ],
    message_handler=message_handler)

# SSE
client = Client(
    "http://localhost:9000/sse", 
    sampling_handler=sampling_handler, 
    log_handler=log_handler,
    roots=[
        "file://home/projects/roots-example/frontend"
    ],
    message_handler=message_handler)

Then run the

uv run python client.py

To run the demo You will need

1 Set SSE Mode

Set both client and server in SSE mode client server

Then run server first

uv run python server.py

2 Two client mode

You can run below for interactive chat mode and see the message exchanged(make sure to be added in the openai group to get api call permission, I added all people in the chat, but ping lipan if see unauthorized)

uv run python client.py chat 

or you can comment out one section by one section and see whats the demo in the code

uv run python client.py