Think about photovoltaics. After all, the photons being converted to useful work are just a form of heat (well, typically pretty high heat.) No reason I can see why photovoltaic technology is not possible in infrared domain.
Similarly, I can imagine a nanodevice using thermal vibrations of some crystal to drive some sort of a crankshaft or some such.
So I don't see why it should be impossible to extract useful work from heat. Of course, in order not to violate the second law you can never extract 100% useful work from the heat; you must always lose some of that input energy (e.g. by scattering it into lower-grade forms of heat that you can't take advantage of within the same design.)
I'm not a physicist (I'm an engineer), but within my limited expertise I don't see any problem with heat-driven devices.