MCPServer (FastMCP) – Ubuntu Setup and GitHub Guide
This repository contains a minimal Model Context Protocol (MCP) server built with FastMCP from the mcp package. It exposes:
Tools:
greet_user(name),add_numbers(a,b),list_files(directory)inserver.pyExample server with resources and prompts in
main.py(add,greeting://{name}, andgreet_userprompt)
The project is configured with a pyproject.toml that depends on mcp[cli] and includes an optional mcp_settings.json for MCP-aware clients.
Requirements
Ubuntu (tested on 22.04+)
Python 3.12+
Git
Optional:
uvfor faster Python dependency management (you can also usepip)
Quickstart (Ubuntu)
Install system prerequisites
Create and activate a virtual environment
3a) Install dependencies with pip
3b) Or install with uv (optional)
Running the server
You have two entry points. Pick the one that matches your use case.
Minimal tool server (
server.py)
This exposes tools:
greet_user(name: str) -> stradd_numbers(a: int, b: int) -> intlist_files(directory: str = ".") -> str
Example server with resources and prompts (
main.py)
This exposes:
Tool:
add(a: int, b: int)Resource:
greeting://{name}Prompt:
greet_user(name: str, style: str = "friendly")
The example runs with transport="streamable-http" (see main.py).
MCP client settings (optional)
If your MCP client supports a settings file (e.g., Windsurf, IDEs), you can point it to your server via mcp_settings.json:
Note: In this repo, mcp_settings.json is configured with an absolute path under this user's home directory. You should update the paths to match your machine if you use it locally.
Project layout
server.py– Minimal FastMCP server exposing three toolsmain.py– Demo FastMCP server with a tool, a resource, and a promptpyproject.toml– Project metadata and dependency onmcp[cli]mcp_settings.json– Example MCP client configuration (absolute paths; edit for your machine).gitignore– Ignoresvenv/and build artifacts
Development tips
Keep your virtual environment out of Git:
.gitignorealready excludes.venvandvenv/.When moving the project to another machine, recreate the venv and install
mcp[cli].If you want to package this repo later, consider adding a proper module and entry points in
pyproject.toml.