Systems & Ops

mcpMarvin

Private

Why this label: It's private work: the service and its code are closed to the public, so there's no link, and the example uses made-up tasks.

Let your AI assistant read today's schedule and add or check off tasks in your Amazing Marvin planner.

mcpMarvin showing an assistant listing today's sample tasks and checking one off

Captured

Current status

Runs as a private service that needs a credential, and the example uses made-up tasks.

Interactive tourNextGen Interactive LLC · Engineering ShowcaseInteractive Architecture Tour
A quick tour of mcpMarvin
Each step shows a real screen filled with made-up data.
mcpMarvin showing an assistant listing today's sample tasks and checking one off
Assistant sessionMade-up data
Assistant session: An assistant lists today's tasks and checks one off.
Assistant session
An assistant lists today's tasks and checks one off.
How it works under the hood: This work is closed to the public, so the write-up covers the stack and what was hard to build.
Try → See → Get: Pick a preset below to update the preview screen and see the exact output.
Where to read morepreset_pills
The showcase page has the full write-up.
What you getWrite-up
Step progress & response previewstep=1/2 · mode=write_up
12
The overview lists the stack and the hardest parts of building it.
Tour step state:action=tour_interstitial_open·step=mcpMarvin-mcpmarvin-session·steward=nextgen_interactive
surface=work_slug_interstitial · saved locally in your browser

Overview

An MCP server that lets an AI assistant read the day's schedule, add tasks and check them off in the Amazing Marvin planner.

Why it was built

Planning apps keep tasks behind their own screens, so an AI assistant can't help plan the day without copy and paste.

Why it helps

Ask the assistant what's due today or to add a task, and the change lands straight in the planner.

Built with

PythonFastMCPhttpxStarlette

What was hard

  • Naming tools and inputs so an AI assistant picks the right one without extra prompting
  • Handling the planner's rate limits and partial failures without losing a task
  • Testing every tool against recorded responses so no test touches a real account

What you can do

  • Reads today's schedule and open tasks
  • Adds, edits and checks off tasks
  • Typed tool definitions an assistant can discover on its own
  • Tests replay recorded planner responses instead of calling the live service