The paper presents an adaptive nonlinear (state-) feedback control structure, where the nonlinearities are implemented as smooth fuzzy mappings defined as rule sets. The fine tuning and adaption of the controller is realized by an indirect adaptive scheme, which modifies the parameters of the fuzzy mapping. The performance index gradients are calculated on-line (using the sensitivity model of the subsystem to be controlled and the controller) to guide the training thus to provide relatively fast learning. Using this scheme an adaptive fuzzy controller design and simulation package was developed with direct link to DSP based real-time implementation. The tool was used to solve highly nonlinear control problems in mechatronics area.