Looks cool. I did something similar 2 years ago in a job, and found it super helpful. My favorite feature was that I added an autocompletion/intellisense engine to the editor window in the browser, and then could start creating requests by just ctrl+space-ing. This worked by parsing all input in a failure tolerant fashion, determining at which field the cursor is, and loading possible matches from the interface definition.

It's pretty cool to interact with an unknown server API in the same fashion than with an unknown library in an IDE.

What I also added was the ability to export requests that one created in the GUI (e.g. as a cURL command including parameters). That made it pretty ease to create requests in the GUI, and then persist and reuse them in other ways.