AI Task Manager

A task manager an AI can operate. Instead of clicking buttons, you tell an AI host something like "add study Python tomorrow, high priority" and it calls a tool on this server to actually do it.

Built with the MCP Python SDK as a small learning project to understand how MCP tools work end to end: natural language in, structured tool call, real data out.

User -> AI -> MCP Task Server -> tasks.json

Features

• Add task - title, priority, optional due date
• List tasks - all tasks with id, priority and status
• Complete task - mark a task done by id
• Delete task - remove a task by id
• Search tasks - find tasks by a keyword in the title
• Update task - change title, priority and/or due date on an existing task

Priorities are low / medium / high. Status is Pending / Completed. Bad input (missing task id, invalid priority, empty title) returns a plain error message instead of crashing.

Project structure

server.py         MCP server and the 6 tools
tasks.json         task storage (plain JSON list)
test_server.py     pytest tests for every tool, including bad input
requirements.txt   dependencies

Install

Requires Python 3.10+.

uv venv
source .venv/bin/activate
uv pip install -r requirements.txt

Run

Use the MCP Inspector to test the server and call its tools directly:

uv run mcp dev server.py

This opens the Inspector in your browser, where you can see and call each tool: add_task, list_tasks, complete_task, delete_task, search_tasks, update_task.

Test

pytest

Example conversation

User: Add study Python, high priority, due Friday.
AI:   Added task 1: Study Python (high)

User: Show my tasks.
AI:   1. Study Python - high - Pending - due Friday

User: Mark it completed.
AI:   Task 1 (Study Python) marked as completed.

Notes

This is intentionally small: one server file, one data file, one test file, six tools. No frontend, no database, no auth. The point was learning how an AI turns a natural-language request into a structured tool call, not building a full product. More on what I learned in LEARNING.md.