I read a short stan handout and it says that the major limitation of stan is it cannot deal with discrete parameters. I wonder what does this exactly mean?

The question I am facing now is like this, for example: I want to estimate the coefficients of a linear regression model Y= X*Beta. I want to introduce an indicator function H to help me decide which coefficients are 0 and which are not 0 (This is like, besides estimating the value of the regression coefficients, I also want to decide which variables I want to keep). So my question is, is stan able to estimate H?

In the process of estimating H, what usually has been done is put a probability on each H[i] (how likely it is to include that predictor), is stan able to give a posterior estimate for that probability?

I am a new user of stan and have a good experience in using this platform. I appreciate peopleâ€™s answer. Thanks.