Protein folding can be described in terms of Langevindynamics. This dynamics can in turn be represented by a “path integral” (analogous to a Feynmannpath integral in quantum mechanics), which is a weighted sum over all paths joining the denatured state to the native state of the protein. In the first part of the talk, we show how one can compute the dominant paths (paths with largest weight) and how one can calculate dynamical quantities (such as rates or transition path times) from these paths. The method and its limitations are illustrated on some simple examples. In a second part of the talk, we show how the Langevindynamics can be modified to obtain a stochastic equation which samples directly and efficiently the transition paths. This new method, called Langevinbridges.