Brms modelling 'simple' correlation in both directions

I’m modelling a correlation between two binary variables while adjusting for confounds, i.e. a phylogenetic co-variance matrix, a spatial co-variance matrix, and an added categorical measure in brms.
However, for the two binary variables, it isn’t really clear which one is the independent/response variable and which one is the dependent/predictor variable. The theory is about their association, it isn’t clear which one impacts the other in a directional or causal sense that would help constructing the formula.
I’ve heard that brms/STAN support models that assess, in the same model/run, both ‘directions’ i.e. variables can take both the independent/response role and the dependent/predictor role. Can anybody give me more info on what such a model is called and how it would be implemented?
Many thanks!