Depression causes clinically significant suffering and/or impairment in the individual's social functioning. There is a consensus in the health care area that, in these cases, it is necessary to offer a comprehensive care model, not restricted to drugs. In this context, there are solutions that can support the diagnosis and interventions for people with a possible depressive profile (PPD), analyzing their behavior on the Internet, more specifically on Online Social Network (RSO). Current research adopts text analysis to try to identify PPD in RSO. However, PPDs can intentionally alter the text to generate a desired social impact. One of the scientific challenges of this project is to combine the textual analysis of posts in RSO with physiological signals and psychometric evaluation scales aiming at a more accurate identification of PPD. In addition to identification, it is understood that the interaction options offered by RSO could be explored as channels of intervention by a computational solution capable of dialoguing with PPD. Thus, this project aims, through a quali-quanti research approach, to investigate a new computing solution for the Internet, which identifies Brazilian users with PPD and to provide and clinically test an autonomous, specialized and personalized intervention via RSO. This solution is as a computational infrastructure and presupposes the construction of a multifactorial model for the identification and an intervention model that explores beyond the text, other media that make sense in RSO as image and music.