Local Task Manager Server

A beginner-friendly Model Context Protocol server built with Python and FastMCP.

This project demonstrates how an MCP server can expose structured tools that allow an AI application to create, read, update, and delete tasks stored in a local JSON file.

Features

• Add a task
• List all tasks
• Filter completed or incomplete tasks
• Mark a task as completed
• Delete a task
• Persist task data in tasks.json
• Return structured MCP responses
• Handle invalid task IDs and input errors

MCP Tools

add_task

Creates a new task.

Inputs:

• title: Required task title
• description: Optional task description

list_tasks

Lists saved tasks.

Optional input:

• completed: Filter by completion status

complete_task

Marks a task as completed.

Input:

• task_id: ID of the task

delete_task

Deletes a task.

Input:

• task_id: ID of the task

Project Structure

task-manager-mcp/
├── main.py
├── tasks.json
├── pyproject.toml
├── uv.lock
└── README.md

Requirements

• Python 3.10 or later
• uv
• Node.js and npm for MCP Inspector

Install Dependencies

uv sync

Run with MCP Inspector

uv run mcp dev main.py

The Inspector should display these tools:

add_task
list_tasks
complete_task
delete_task

Example Task

{
  "id": 1,
  "title": "Prepare Day 5 MCP article",
  "description": "Connect the Task Manager MCP Server to Trello",
  "completed": false
}

Example Flow

User request
    ↓
MCP client
    ↓
Local Task Manager MCP Server
    ↓
Python tool
    ↓
tasks.json
    ↓
Structured response

What This Project Teaches

• Creating an MCP server with FastMCP
• Exposing Python functions using @mcp.tool()
• Generating tool schemas from type hints
• Returning structured data
• Validating tool inputs
• Handling expected errors
• Persisting data in a local JSON file
• Testing MCP tools with MCP Inspector

Roadmap

The next version will connect the same task-management capabilities to Trello:

add_task → Create a Trello card
list_tasks → List Trello cards
complete_task → Move a card to Done
delete_task → Archive a Trello card

Author

Akbar Shaik

Machine Learning Engineer teaching MCP servers and AI agents from scratch.