Rather annoying, but typically your fragment file will have the barcodes in column 4.

chr1    3003761 3003928 TTTGCGACAGTAGGTG-1      48
chr1    3008963 3009663 AAGGTGCAGTCAATTG-1      3
chr1    3012286 3013245 AACCTAATCCCGTTGT-1      2

You can combine awk and bedtools via something like

zcat fragments.tsv.gz | awk '{print > $4".tsv"}'
for fn in `ls *.tsv`; do
    base="${fn%%.*}"
    bedtools genomecov -bg -g /path/to/ref.fa.contigs -i "${fn}" > "${base}.bdg"
done

which will split the fragments into per-cell tsv files, and convert each to bedgraphs. You can use gnu parallel to speed up the process (after splitting the fragment file). Bedops might be able to do this all in one go; but I didn't find an option the last time I looked.