Advent of AI 2025 - Day 17: Building a Wishlist App with Goose and MCP-UI

The Advent of AI series utilizes Goose, an open source AI agent that automates engineering tasks and goes beyond code suggestions. Goose is a local, extensible, and open source AI agent that can build entire projects from scratch, write and execute code, debug failures, and interact with external APIs. For Day 17 of the Advent of AI challenge, the author built a Winter Wishlist app using MCP-UI, which allows users to add wishes, view them in a nicely formatted UI, grant wishes when they come true, and delete wishes they no longer want. The app is built on top of a ChatGPT app TypeScript template, which made spinning up the wishlist server much faster. The implementation stores wishes in memory based on MCP session ID, but a more robust solution would be needed for production. The author encountered an issue with iframe sizing in Goose, but was able to resolve it using a ResizeObserver and posting messages to the parent frame. The key learnings from this experience include the power of MCP UI for creating visual interfaces and the importance of starting with a solid template. The author encourages readers to join the Advent of AI challenge, even if they missed earlier days, and to stay in touch through their social media channels. The development workflow for this project was also highlighted, with the ability to build and edit components in Storybook and see them update in Goose in real-time. Overall, the project demonstrates the potential of Goose and MCP UI for building innovative and interactive applications. The author's experience with the project showcases the benefits of using open source tools and templates to speed up development and create more efficient workflows.