Django itself works with an HTTP Request/Response cycle. “Something” (browser, program, app, etc) issues an HTTP request to your server. Django receives that request and returns a response. That response can be pretty much anything - HTML, JSON, text, PDF, whatever.
So if MIT Appinventor allows you to issue HTTP requests to a server and do something useful with the response, then yes, they’ll work quite well together.
I would love to see this, when you get it running.
One thought I have, how will you test when you are in development with Django locally? I am asking because my first thought is since MIT AI is on the web, the request will come from their ip to your local machine. How is that done?