The inversion of magnetic data to characterize dyke swarms is a challenging problem due to complexity of the anomalies arisen from multiple sources in different depth levels and with variable magnetic properties. It makes difficult to account for the number of dyke unities along a profile as well as unambiguous estimates about position, depth to the top and magnetization properties for single unities.We develop a procedure to obtain mean properties for juxtaposed dykes by using non-linear data inversion for thin-sheet models, with subsequent cluster analysis of the obtained solutions. This procedure is implemented in Python by program INVDYKE according a sequential inversion approach, initially inverting a nonsensitive quantity to the magnetization direction and then the magnetic anomaly itself. INVDYKE is validated with numerical simulations and then applied to real data sets from the dyke swarm of the Ponta Grossa Arch to identify mean parameters describing the associated magnetic sources.