Along with increased complexity of the models used for scientific activities and engineering, come diverse and greater uncertainties. Today, effectively quantifying the uncertainties contained in a model appears to be more important than ever. Scientific fellows know how serious it is to calibrate their model in a robust way, and decision-makers describe how critical it is to keep the best effort to reduce the uncertainties about the model. Effectively accessing the uncertainties about the model requires mastering all the tasks involved in the numerical experiments, from optimizing the experimental design to managing the very time consuming aspect of model simulation and choosing the adequate indicators and analysis methods. In this paper, we present an open framework for organizing the complexity associated with numerical model simulation and analyses. Named mtk (Mexico Toolkit), the developed system aims at providing practitioners from different disciplines with a systematic and easy way to compare and to find the best method to effectively uncover and quantify the uncertainties contained in the model and further to evaluate their impact on the performance of the model. Such requirements imply that the system must be generic, universal, homogeneous, and extensible. This paper discusses such an implementation using the R scientific computing platform and demonstrates its functionalities with examples from agricultural modeling. The package mtk is of general purpose and easy to extend. Numerous methods are already available in the actual release version, including Fast, Sobol, Morris, Basic Monte-Carlo, Regression, LHS (Latin Hypercube Sampling), PLMM (Polynomial Linear metamodel). Most of them are compiled from available R packages with extension tools delivered by package mtk. mtk: A General-Purpose and Extensible R Environment for Uncertainty and Sensitivity Analyses of Numerical Experiments