something like this?

 r = RandomVariate[BinormalDistribution[.5], 100];
 hg = Histogram3D[r]

![enter image description here][1]

 hl = HistogramList[r];
 Show[hg, Graphics3D[{Thick, Red, 
 Map[Append[#, 5] & , First@Cases[Normal@First@
 Cases[ListContourPlot[ 
 Flatten[Table[{
 Mean@hl[[1, 1, j ;; j + 1]], 
 Mean@hl[[1, 2, i ;; i + 1]],
 hl[[2, j, i]]},
 {i, Length@hl[[1, 2]] - 1}, {j, Length@hl[[1, 1]] - 1}], 1], 
 Contours -> {5}], _GraphicsComplex, Infinity],
 _Line, Infinity], {2}]}]]

![enter image description here][2]

By the way, It would be cleaner to work with `SmoothDensityHistogram`, but I can't figure how to coax it to give a single contour line at a specified level..


 [1]: https://i.sstatic.net/noJT9.png
 [2]: https://i.sstatic.net/ehlsB.png