# AI MCP Tools

A lightweight local MCP server demo built in Python.

This repository contains a small set of tool definitions for `mcp`, showing how to expose simple operations through a local MCP server and manage the environment with `uv`.

## What this repo contains

- `main.py`: a minimal MCP server with five tools:
  - `add(a, b)`
  - `subtract(a, b)`
  - `multiply(a, b)`
  - `divide(a, b)`
  - `long_process(steps)`
- `MCP.md`: notes on MCP architecture and the main components of the protocol.
- `pyproject.toml`, `uv.lock`, and `.venv/`: the source configuration and environment state created through `uv`.

## Why this setup

I used `uv` to initialize and manage the Python environment, keep dependencies locked, and keep the repo light and reproducible. The goal is a clean, easy-to-run demo rather than a copied template.

## Getting started

1. Create or activate the virtual environment:

   ```powershell
   uv venv .venv
   .\.venv\Scripts\Activate.ps1
   ```

2. Install dependencies:

   ```powershell
   uv install
   ```

3. Run the MCP server locally:

   ```powershell
   mcp dev main.py
   ```

## Notes

- The repository is designed for local development and experimentation with MCP tools.
- `uv.lock` stores the exact dependency versions, so the environment can be reproduced.
- `.gitignore` excludes the virtual environment and Python cache files.

## Quick commands

- `uv init` — initialize the project config with `uv` (already done)
- `uv install` — install packages and update the lock file
- `uv venv .venv` — create the local virtual environment
- `mcp dev main.py` — start the MCP tool server