As the data are available as a VCF file, you can just use the custom annotations feature of the VEP to pull in the sites and any frequency data you want:
curl -O ftp://ftp.broadinstitute.org/pub/ExAC_release/release0.3/ExAC.r0.3.sites.vep.vcf.gz
curl -O ftp://ftp.broadinstitute.org/pub/ExAC_release/release0.3/ExAC.r0.3.sites.vep.vcf.gz.tbi
perl variant_effect_predictor.pl -i example_GRCh37.vcf -cache -custom ExAC.r0.3.sites.vep.vcf.gz,ExAC,vcf,exact,0,AF
Here I've added the AF field (it will appear as ExAC_AF in the VEP output), but you could add any of the fields in the INFO block of the VCF by listing them at the end of the --custom string. See http://www.ensembl.org/info/docs/tools/vep/script/vep_custom.html for more info.
You could get a list of fields by doing (this gets the VCF header)
tabix -h ExAC.r0.3.sites.vep.vcf.gz 1:1-1
There's a caveat here in that the VCF carries the per-population allele counts but not the frequencies, so to derive the frequencies you'd need to do a little math(s). It's probably worth writing a plugin to deal with these (and a very simple one it would be too, see the dbNSFP plugin http://www.ensembl.info/ecode/dbnsfp/ for an example of one that uses tabix to retrieve data from a file), but in the short term the above solution should at least get you part of the way.