Did not used it myself yet, but there is a new program claiming to be better than primer3:

See: http://nar.oxfordjournals.org/cgi/content/abstract/37/13/e95

Download it from: http://pythia.sourceforge.net

I know that many people use the primer3 software to design primers. There are also two web-interfaces where you can try out the parameters, and there are many of them.

The program eprimer3 is part of the EMBOSS tools, maybe the best way of getting a command-line executable of primer3.

I didn't see a mismatch parameter though, because for a 'real' PCR-primer to allow for mismatches is a bad idea. I noted the '' in your question, so you might not really be looking for primers. So what is it then? If this is just a piece of sequence then you do not need to bother with uniqueness or melting temperature constraints.

[EDIT] Just saw this today which could be another piece of software to check out. I didn't read it carefully, but it looks like it builds on primer3. [/EDIT]

Since you have so few pairs, you could just run them through a read-aligner allowing mismatches, and find pairs that are aligned within X basepairs in the reference genome.

Actually, you could probably even use the paired-end feature of most aligners so that the aligner only find the pairs with nearby matches. For either of those methods, you could use bowtie.

As far as I know, primer3 does allow gaps and mismatches, so that may also be an option if you just want to see if you have good primers, but you'd probably have to use the command-line version.

or, you could use BLAST and find nearby hits between the pairs.

You could try e-PCR or in-silico PCR. e-PCR is available to run locally.

MFEprimer has an command-line to check the specificity of primers, you can download it from the web site: http://biocompute.bmi.ac.cn/MFEprimer/

(better late than never)

One can use blastn with -task blast-short, and it's quite fast. As explained in the manual (section BLAST+ features), "blastn-short is BLASTN program optimized for sequences shorter than 50 bases". Further down (table C2), it says "blastn-short is optimized for sequences less than 30 nucleotides" and specifies word_size=7 (versus 11 for blastn) and reward=1 (versus 2 for blastn), all the other parameters remaining as for blastn.

In a Unix-like machine, with NCBI BLAST 2.2.26+:

zcat genome.fa.gz | formatdb -i stdin -l formatdb.log -p F -n genome
blastn -query primers.fa -db genome -task blastn-short -out primers_one_genome.txt -outfmt 7

If one knows both primers around the sequence they amplify, I recommend concatenating them into a single sequence, and map it to the genome. BLASTn-short will usually (likely?) return two best hits, one per primer, separated by a gap on the genome.