A new crystal plasticity model is presented. The model allows to compute a solution to the mechanical state of polycrystals taking into account the topology of the microstructure without building a representative volume element in three dimensions. Instead, topological information is directly extracted from heterogeneous experimental data, which might be acquired using different common techniques. The model is formulated using a mean-field approach similar to the one used by the viscoplastic self-consistent model of Lebensohn and Tomé, and solved using the gradient descent method. However, at difference of the VPSC model, where grains are assumed to be elipsoidal inclussions in a homogeneus medium, different interaction schemes can be defined between homogenisation units (not necessarily grains) and their environments. These homogenisation units are assumed to be the superposition of homogeneous finite elements. These elements can be chosen to belong, for instance, to different phases, grains with certain properties (such as phase and orientation, and also size and shape), individual grains, or even individual data points (like in a full-field model). For each homogenisation unit, an environment is defined, which will vary depending on the topology of the microstructure. Computational requirements will depend on the number of homogenisation units under consideration, which can be chosen independently of the amount and resolution of experimental data, and can work at several levels. For example, homogenisation units could be grains with a certain phase and orientation, but it is also possible to obtain a more refined solution for the individual points of a particular set of grains. In addition to demostrating how the model works through several practical examples, its implementation in the APL programming language is presented. Using an array programming approach, the premises of the model can directly be translated into runnable code, resulting in a simple but efficient implementation.