Alternatively, if starting from nucleotide sequence, one would do a blastN search of the NCBI nr database, or the dbEST database directly. If starting from a gene name, use NCBI nucleotide Entrez to find a RefSeq entry (one with NM_xxxxx for an accession), and click on <Link> to get a popup menu and look for Unigene. That will give a list of available cDNAs (called ESTs) with some annotation about whether any are thought to be full length.
Only about 12,000 human genes are currently listed as represented by full length cDNAs (3/2003). If none is available, you may be able to identify 2 EST clones that might be recombined to a full length cDNA. Try a BlastN search of the full sequence (from the NM_xxxx entry) versus dbEST in multiple alignment view to visualize this. If parts of the gene are not cloned, then you'll have to get the longest one you can and do RACE (a PCR method to recover the rest of the cDNA), or some other method to get the rest of the clone.
At NCBI, references to IMAGE clones means that the cDNA can be retrieved
from the I.M.A.G.E consortium (see below).
I.M.A.G.E also has tools to visualize how the various cDNA clones overlap.
An MGC designation means information is also available at the Mammalian
Gene Collection page.
Here is a tutorial navigating the I.M.A.G.E. system that was written in 2001 and may be somewhat dated now.
You may also find specialized cDNA banks by searching BioHunt or a general search engine. E.g. cDNA Arabidopsis.
Remember that "full length" to these folks means the initiator AUG is
present. It doesn't necessary mean that the 5' UTR is all present.
They could have also misidentified the initiator AUG and be missing part
of the coding sequence. Also remember that cDNAs often have mutations
or frame shifts due to reverse transcription error, and EST clones are
often sequenced in a low reliability mode. Finally, remember that
alternative splicing may cause there to be more than one valid "full length"
configuration.