The gridlib project formulates the foundation for further projects. Prime principle is the idea of a framework - finding a frame in which tasks of numerical simulation can be solved. In particular it shall allow for the steadily growing amounts of data and, by showing a clear concept, for the integration of variuos simulation, visualization and rendering algorithms. To fulfill large projects” needs of technical and algorithmic maintainance, usability in teaching and research and integration of third-party modules, the gridlib is designed in a fully object oriented way. This encapsulation of algorithms and data also assists with the transparent use of parallel programming concepts, modular structure and clean interfaces for external subsystems with a procedural interface. The gridlib project adopts a hybrid, adaptive, hierarchical unstructured approach to grids. The reachable performance of these algorithms in the domain of grid management and visualization can be combined with the numerical performance of procedural systems. The object oriented high-level concept enables a very flexible use of algorithmic optimizations for the special cases.