My First MCP: No More Manual Bug Tickets

This week I shipped my first MCP, and it solved a problem I was tired of pretending was small.

I did not want to manually create bug tickets anymore.

When I am playtesting and something breaks, context dies fast. If I stop, alt-tab, open Trello, make a card, type details, upload an image, and try to remember exact coordinates and state, I lose momentum and often lose accuracy too.

That was the loop I wanted to kill.

Now I hit F9 in game, type what happened, and submit. The report includes the useful context at the moment of failure, plus screenshot support, and lands as a Trello ticket. The workflow changed from “I should file this later” to “it is already filed.”

The interesting part is that the biggest win was not complicated AI logic. It was setting a clean boundary.

The game submits one bug-report intent. The MCP layer owns Trello details, auth, and tool behavior. That gave me a shared contract I can use from both runtime tools and agent workflows without rewriting the same request handling over and over.

I wrote the architecture tradeoff in an earlier post, Should I Build My Own MCP for Trello? Yes, But Keep It Thin. This post is the result after shipping it: the boring manual step is gone, and my iteration loop is faster.

For me, that is the real test for any workflow tool. Not “is it clever,” but “does it remove friction every single day?”

This one does.