In this submission, I’m showing how any AI agent or harness connected to Paper through MCP can automate the entire design process. I’m using a streetwear brand project called BUZZTAIL to demonstrate a workflow that bridges data, logic, and visual design.
The process starts in Notion, where I’ve prepared everything from product descriptions to extreme test scenarios to see how durable the design will be. Using Cursor as the AI agent, I input instructions to fetch that data and build the UI structure in
Paper.design in real-time through the MCP integration.
After the initial build, I refined the visual by uploading my custom hero illustration and using Paper’s built-in AI to generate and replace product images.
The highlight of this project is the UI Stress Test. I instructed the agent to inject "messy" data from Notion including extremely long text and foreign characters into the finalized design. The result shows that the layout remains solid and doesn't break, proving that the system is completely data-proof. This workflow eliminates manual design debt and ensures that the UI stays consistent no matter what content is thrown at it.
You can check out the full design on Paper here:
https://app.paper.design/file/01KPP5BHT9PFEP7GHAX258ZPJV/1-0
And if you want to see the "messy" data I used for the stress test, the Notion database is here:
https://www.notion.so/348d6d1c0ade805180f1f5bfc4946887?v=348d6d1c0ade8061a24e000ce345ea7e&source=copy_link