If you only have to make a few access, you can skip indexing. Let's assume that you downloaded RefSeq and uncompressed it in a file called human.rna.gbff in your current directory, the command for retrieving the sequence of NM_004348 in FASTA format is seqret human.rna.gbff:NM_004348. Here is an example using the -filter option to send the result to the standard output.

$ seqret -filter human.rna.gbff:NM_004348
>NM_004348 NM_004348 Homo sapiens runt-related transcription factor 2 (RUNX2), transcript variant 3, mRNA.
ggactccaactgggagggggcgccagcgcacccaggacgcggtgccccaagggaccccat
tccaagctgcaaactcaagtcccccgcctccccaggccaaacacacccccgcgcccgttc
gcagctgccaccccgggtgttccaaagactccggcaaagatctgcgcctcccgggtctcc
ctacccgccagcccgactgcgccgcatcgccagactctgttggcccatcagactccgccc
ggcagtcggcctcatcaaacttgatttctcacctcctcagccccatcacctccatcctct
ttccccccgtctcgccttcacccccccaatttcctccttgcccctcatttccaccctcct
ccccctcccccggccacttcgctaacttgtggctgttgtgatgcgtattcccgtagatcc
gagcaccagccggcgcttcagccccccctccagcagcctgcagcccggcaaaatgagcga
cgtgagcccggtggtggctgcgcaacagcagcagcaacagcagcagcagcaacagcagca
(shortened for brievty)

If you need to to access repeatedly to RefSeq records from arbitrary directories, have a look at the manual of the dbxflat command.

RefSeq is available for download from the NCBI's FTP service.