There is also bedtools getfasta. It uses Erik's code under the hood.

$ cat test.fa
>chr1
AAAAAAAACCCCCCCCCCCCCGCTACTGGGGGGGGGGGGGGGGGG

$ cat test.bed
chr1 5 10

$ bedtools getfasta -fi test.fa -bed test.bed -fo test.fa.out

$ cat test.fa.out
>chr1:5-10
AAACC

Docs: http://bedtools.readthedocs.org/en/latest/content/tools/getfasta.html

And it is wrapped in pybedtools as well: http://pythonhosted.org/pybedtools/autodocs/pybedtools.BedTool.sequence.html?highlight=fasta

I think samtools faidx feature exactly does what you need. If you give a reference genome and coordinates as input it will extract the sequence. Check samtools See below:

faidx samtools faidx <ref.fasta> [region1 [...]]
Index reference sequence in the FASTA format or extract subsequence from indexed reference sequence. If no region is specified, faidx will index the file and create <ref.fasta>.fai on the disk. If regions are speficified, the subsequences will be retrieved and printed to stdout in the FASTA format.

You will have to first create the fasta indexes of the reference genome fasta file and then use this command.

This is a shameless plug, but I just uploaded a pure python implementation of faidx to GitHub. I'm calling it pyfaidx. It will also work for you. You might also check out:

1. FastaHack
2. brentp's fastahack-python, which inspired my implementation

Both of which require compiled C code, but aren't actually much faster than the pure python version.