Doing this with 10X data is a pretty large challenge due to the data sparsity. Most genes may only have 1 or 2 reads aligning, and coverage across the gene will obviously be biased. This paper has some good information. In general, you will miss much, along with all the other caveats that come with variant calling in RNA-seq data.

There are ways to look at known variants (e.g. from WGS/WES on the same sample) in your scRNA-seq data, which can be useful when comparing mutation frequencies across groups/types of cells (malignant vs normal, etc). vartrix is the 10X tool made for this, and it's pretty easy to use.