Dear Jose,
I think the best strategy for you is to try to follow this tutorial:
As you will see, this tutorial takes in an TAB-delimited file that shows the distribution of taxa across metagenomes. You can imagine that the same steps can show you the distribution of gene clusters across genomes. You must be getting an output from Roary that gives you that data.
Of course this wouldn't allow you to do a lot of things you can do with native anvi'o pangenomes. But if you do this and show us that there is promise, we may be able to implement something that imports gene clusters from Roary outputs to do the rest of the workflow natively. I know many people like Roary and we would be happy to modify anvi'o to support Roary so anyone can have access to the best of both workflows.
Best wishes,