Part 1: Connecting AI to My Daily Tools – A Simple MCP-Powered TODO App

Imagine asking an AI to manage your TODO list, run a local script, or update a file - it does exactly that, like a virtual assistant plugged into your computer. This is now possible thanks to something called MCP (Model Context Protocol). 

What is MCP? 

MCP (Model Context Protocol) is a lightweight protocol proposed by Anthropic that enables AI models to interact with tools in a structured and transparent manner. It lets you define tools in an AI-readable format — including their names, parameters, descriptions, and return types — so that LLMs like Claude, GPT (and others) can understand and call them safely. With MCP, you can turn your scripts, utilities, and services into AI-callable tools, effectively giving the model "extensions" into your local environment or applications. 

Image Source: fb.com/devtalks

My First Experiment: A TODO List MCP Server 

To explore MCP in action, I created a simple MCP server that exposes two tools: 

 📝 addTodo(day, task) — Adds a task under a specific day in a TODO list. 

 📖 readTodoList() — Reads and returns the current TODO list. 

The TODOs are saved in a plain todo-list.txt file, stored wherever the user prefers (customizable via config). 

I integrated this server with Visual Studio Code’s AI Agent Mode, and it worked like magic: 

 “Add ‘Prepare slides for meeting’ to Thursday’s TODO list.” 

“What’s on my TODO list?” 

 The AI agent calls the tool via MCP, and updates the file instantly. 

Explore more from this repo: https://github.com/mukitul/mcp-server-test-project

Why This Matters 

This is just a simple example, but it shows how MCP can unlock powerful, personalized workflows. You can use the power of AI LLM models to 

- Automate parts of your local workflow 

- Trigger shell scripts, analyze files, or control apps in natural language. 

Not just about your workflow, your application's user can interact with natural language to get result from the application. 

MCP brings a layer of tool-augmented intelligence to your everyday computing. 

What’s Coming in Part 2 

In the next post, I’ll go a step further: 

 - Set up a locally running LLM (no internet or cloud dependency) 

 - Build an MCP client that communicates with the TODO server 

 - Create a fully local, private, LLM-powered system that understands and executes tasks