Skip to case study

Oct 2026

Dayfold: learning MCP through a shared todo app

A simple todo app for experimenting with an MCP server, agent authentication and user-controlled access to shared task lists.

MCP · WorkOS AuthKit · WorkOS Connect

Visit project

Problem

I wanted to build an MCP server and explore how AI agents could use it to manage tasks through automations and scheduled workflows. A todo app gave me a simple, practical place to learn.

My role

I built Dayfold and its MCP server so people and agents could work on the same shared task lists. Agents can create, reschedule, complete and reopen tasks through MCP.

Decision

I used WorkOS AuthKit for user sign-in and WorkOS Connect for agent authentication. Users choose which projects each agent can access, give it read-only or edit permissions, and revoke access anytime.

Result

Dayfold is available to try as a shared task list for people and agents. Unfinished tasks carry over to today based on the project's time zone. I kept the task list simple on purpose so I could focus on giving agents useful access while keeping the user in control.

Further reading

This account describes my own MCP and AI agent experiment. Try the live app using the project link above.