A native iOS visual workflow for Qwen Code: edit SwiftUI directly from the running app

[ivanzhaowy]

While using Qwen Code for native iOS projects, I kept running into a disconnected part of the workflow: agents are good at reading and modifying code, but the UI state I am looking at in the Simulator often has to be translated back into screenshots and a long text description.

That is why I built Monad Design. It is not another app generator. It adds a visual implementation layer to existing coding agents.

The idea is simple: The app is the canvas.

You select UI directly in the running iOS app and describe what you want to change. Qwen Code continues to understand the existing project and modify the real SwiftUI source code. After rebuilding, you return to the Simulator, compare the result, and decide whether to accept it.

I have completed the local Qwen Code integration setup:

• The MCP configuration is written to ~/.qwen/settings.json
• The Monad Design skill is installed at ~/.qwen/skills/monad-design
• The installer supports both project and global scopes
• Existing Qwen Code configuration is preserved rather than overwritten wholesale

Installation:

npx monad-design install

One important boundary: this post describes an installation and configuration path that I have verified. I have not yet completed a full Qwen Code end-to-end run of “select UI → edit source → rebuild → compare,” so I am not presenting it as a fully verified integration demo. Once that test is complete, I will add the configuration, source diff, and Simulator results.

If you use Qwen Code for SwiftUI work, you can try Monad Design. Instead of translating the interface back into a prompt, I want the agent and developer to work around the same running app.

Project: https://github.com/Monadix-AI/monad-design

Disclosure: I am the developer of Monad Design, and the project is open source. I would also love to hear how others currently provide runtime visual feedback to Qwen Code.