Binary packages are available for:
Official Debian package (python-mvpa2). Backports for Debian and Ubuntu releases are available from the NeuroDebian project.
Python 2.6 is available from the github download area. Also installers of PyMVPA for Python 2.5, 2.6 and 2.7 are available from the Unofficial Windows binaries collection of Christoph Gohlke.
If there are no binary packages for your particular operating system or platform, you have two opportunities. You could compile PyMVPA (which is not strictly necessary anyways unless you need to use SVM or SMLR). The manual contains instructions to build PyMVPA in various environments (Windows, Fedora, OpenSuse, ...). Alternatively you could use NeuroDebian Virtual Appliance which allows in a matter of minutes to get a full featured (Neuro)Debian GNU/Linux distribution running on top of your system, thus providing easy access to over 30,000 software products many of which are related to machine learning and neuroscience.
Source code tarballs of PyMVPA releases are available from the download area. Alternatively, one can also download a tarball of the latest development snapshot (i.e. the current state of the master branch of the PyMVPA source code repository).
To get access to both the full PyMVPA history and the latest development code, the PyMVPA Git repository is publicly available. To view the repository, please point your webbrowser to gitweb: http://github.com/PyMVPA/PyMVPA
To clone (aka checkout) the PyMVPA repository simply do:
git clone git://github.com/PyMVPA/PyMVPA.git
After a short while you will have a PyMVPA directory below your current working directory, that contains the PyMVPA repository.
More detailed instructions on installation requirements and on how to build PyMVPA from source are provided in the manual.