Hi @Sara_Jeanne08!
Unfortunately q2-cutadapt probably won't work for you --- it only supports demultiplexing reads with the barcode still in the read sequence (I think this is what you are talking about when you said that your reads need to be "joined" together).
This option could get you moving quickest - check out this Community Tutorial for some help on getting started with that route.
Another option would be to process these data as single-end reads - just follow the EMP import and demux protocol, but that assumes you can demux forward reads with just the forward barcodes (e.g., this isn't a dual-indexing scenario). Similarly, if this isn't a dual-indexing scenario, you could import and demux your forward reads, import and demux your reverse reads, export both resulting sets of demuxed reads, and re-import as paired-end demux data using a manifest format. That is a lot packed in there, and I might be missing some critical assumption, but if that is something you want to pursue we could try and walk you through it.
Those options make a lot of assumptions, so your mileage may vary. In the meantime, we will track this use-case for future support of this sequence format. Thanks, and if you get stuck, you know where to find us! 