You have 2 problems to solve here: fetching the sequences and parsing them to extract CDS.

To deal with the second problem: install Bioperl, if you have not already done so. Then, take a look at the SeqIO how-to. If you installed the accessory scripts, there's a handy utility named bp_extract_feature_seq, which you can run like this:

bp_extract_feature_seq -i NC_005213.gb --format genbank --feature=CDS -o NC_005213.fa

It will write a fasta file containing the coding sequences of all CDS features.

You'll want to automate the process of fetching sequences by looping through the replicon accessions. Here's some sample code from the Bioperl tutorial which will fetch a sequence from RefSeq and write it in GenBank format:

#!/usr/bin/perl

use strict;
use Bio::Perl;

my $seq_object = get_sequence('refseq', "NC_005213");
write_sequence(">NC_005213.gb", 'genbank', $seq_object);

It should not be too hard to write a loop into that, using the NC_* accessions from your file.

An addendum to Jorge Amigo's answer. Use BioMart as he describes. Additionally, set the gene type to be 'protein_coding'. Then click the 'Perl' button on the top right. Edit the script that it returns to include a loop over your set of organisms. e.g.

$query->setDataset("<my organism>");
$query->addFilter("refseq_peptide", ["<id 1>","<id 2>"]);
$query->addFilter("biotype", ["protein_coding"]);

You may have to normalise your IDs so that they are all of the same type (e.g. Refseq)

you can use BioMart to retrieve such information. I would select the latest version of the Ensembl Genes database (currently #59) and then select the appropriate organism. you will then only have to enter the list of IDs you have, using the "ID list limit" filter and selecting from the combo box the type of IDs you are going to enter for that particular organism. you will have to do it organism by organism though.

From the NCBI, you can use ESearch and/or EFetch to retrieve an XML description of your record (search biostar for some examples).

Here is an example.

Then, using XSLT/XPATH, you can extract the CDS (again, search biostar for some examples).

XPATH example:

/GBSet/GBSeq/GBSeq_feature-table/GBFeature/GBFeature_quals/GBQualifier[GBQualifier_name='translation']/GBQualifier_value

but I'm afraid the annotation of the NCBI are not homogeneous, so the word 'translation' might not be the only one to match the CDS.