Skip to content

VSCode MCP Server

DevOps / vscode

MCP (Model Context Protocol) is an open-source standard for connecting AI applications to external systems.

MCP Concept

  • Resources: File-like data that can be read by clients (like API responses or file contents)
  • Tools: Functions that can be called by the LLM (with user approval)
  • Prompts: Pre-written templates that help users accomplish specific tasks

Simple MCP Server demo

using python mcp library

uv

Must of the example use uv the new package installer and runner I still use pip and the old virtual environment

using logging

Using logging and not print

1
2
3
4
5
6
7
8
import logging


logging.basicConfig(level=logging.INFO,
    format="%(asctime)s [%(levelname)s] %(message)s"
)

log = logging.getLogger(__name__)

Logging

The logging from the mcp server mark is warning and from [server stderr] as source , is ok and not an error the mcp output to stderr we can show from the image below that the rest of the logging line describe to output from our mcp server

alt text

Install

1
2
pip install mcp[cli]
pip install trio #  friendly Python library for async concurrency and I/O

Very simple server

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
from mcp.server.fastmcp import FastMCP
import logging


logging.basicConfig(level=logging.INFO,
    format="%(asctime)s [%(levelname)s] %(message)s"
    )

log = logging.getLogger(__name__)


# Create an MCP server
mcp = FastMCP("Demo")
log.info("------- Starting MCP server -------")


@mcp.tool()
def add(a: int, b: int) -> int:
    """Add two numbers together."""
    return a + b

# Prompt: greeting
@mcp.prompt()
def greet(name: str) -> str:
    """Generate a greeting prompt."""
    return f"Please write a short greeting for {name}."

# Resource: static greeting
@mcp.resource("hello://{name}")
def hello(name: str) -> str:
    """Return a static greeting resource."""
    return f"Hello, {name}!"

# Entrypoint
if __name__ == "__main__":
    mcp.run()

Register the server in vscode

From mcp.json file we can manage the server: start, stop, show output(log) and more

mcp.json
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
{
    "servers": {
        "demo-mcp": {
            "command": ".venv/bin/python",
            "args": [
                "mcp_demo.py",
                "stdio"
            ],
            "type": "stdio"
        }
    }
}

The tools relevant only from agent mode and can config using the tool icon

alt text

Manage

From vscode extensions tab

alt text

Usage from copilot

@mcp.tool()

  • Purpose: Expose a Python function as a structured action the client (Copilot Chat, Cursor, Windsurf, etc.) can call.
  • Input/Output: JSON-serializable arguments and return values.
  • Use case: Do something deterministic (math, file ops, API calls).

permission

agent mode
1
Use the add tool with a=2 and b=3

alt text

@mcp.prompt()

  • Purpose: Expose a prompt template (reusable instruction).
  • Input: Arguments you want to inject into the prompt string.
  • Output: A text template that will be fed to the AI.
  • Use case: Style instructions, boilerplate phrasing, workflows.
1
Use the greet prompt with name="Amir"

@mcp.resource()

  • Purpose: Expose structured data sources the client can browse/fetch.
  • Input: Usually a URI-like parameter (you define the pattern).
  • Output: String or JSON content.
  • Use case: Provide documents, snippets, templates, or data.
1
Fetch hello://Alice

More Demos

Generate files from templates

Resources