Novice users usually find it hard to manipulate models by using traditional Model-Driven Development techniques, because of the gap between the modeling tools and these users’ mental models. In this context, multi-touch interfaces emerge as an alternative to make it easier for novice users to interact with the models by using natural gestures and taking advantage from the popularity that touch-based devices have achieved. In this paper, a metamodel infrastructure and a set of heuristics are presented to automatically generate multi-touch visual editors for manipulating models. The editor generated is driven by a metamodel that also prevents the user from creating not valid models. These heuristics have been validated while developing an environment for novice users, such as psychologists or physiotherapists, for the treatment of people with Acquired Brain Injury.