In this contribution the software design and implementation of an analysis server for the computation of failure probabilities in structural engineering is presented. The structures considered are described in terms of an equivalent Finite Element model, the stochastic properties, like e.g. the scatter of the material behavior or the incoming load, are represented using suitable random variables. Within the software framework, a Client-Server-Architecture has been implemented, employing the middleware CORBA for the communication between the distributed modules. The analysis server offers the possibility to compute failure probabilities for stochastically defined structures. Therefore, several different approximation (FORM, SORM) and simulation methods (Monte Carlo Simulation and Importance Sampling) have been implemented. This paper closes in showing several examples computed on the analysis server.