It seems as though I've got complete-control over the

app

server, and none at all, over the

db

server, is that correct?

Yes, the architecture looks like you’ve described above. There is a Prisma GraphQL application that acts as a thin layer over your database. That API provides CRUD-based query and mutation capabilities. This application can, but doesn’t has to, be deployed on Graphcool servers. You can deploy it on a separate cloud provider as well (e.g. Digital Ocean). The CRUD functionality gets scaffolded from your data model. The second application is the actual front-facing GraphQL API that would hold your business logic and uses Prisma for manipulating / reading your data.

@andre Are there any guides/tools to help with deployment on AWS say?

Not sure. I think that the only cloud environment which is officially supported for the moment is Digital Ocean. There is an open issue for other environments though: https://github.com/graphcool/prisma/issues/1350