You can try the e-utilities, for example e-fetch.

A greedy solution could be to get the genbank file corresponding to your gi with efetch with get, something like that in Perl:

$gbk = get(http://eutils.ncbi.nlm.nih.gov/entrez/eutils/efetch.fcgi?db=nucleotide&id=$id&retmode=text&rettype=gb);

Then parse the genbank with BioPerl:

my $gbk_stream = Bio::SeqIO->new(    -file => $gbk,
                                    -format => 'GenBank');
my $seq_obj = $gbk_stream->next_seq();

if (defined ($seq_obj)){
  for my $feat_object ($seq_obj->get_SeqFeatures) {
    if ($feat_object->has_tag('cds')) {
      $cds = $feat_object->get_tag_values('cds'));
    }
  }
}

If you only need the CDS it may be a little bit to much. But if you are going to need other information from the genbank file it could be useful.

There is probably a simpler and more elegant solution.