A distribution static compensator (D-STATCOM) is a custom power device connected in parallel to a power system to address electric power quality problems caused by reactive power and harmonics. To obtain high performance from a D-STATCOM, the D-STATCOM's dq-axis currents must be controlled in an internal control loop. However, control of the D-STATCOM's currents is difficult because of its nonlinear structure, cross-coupling effect between the dand q-axis, undefined dynamics, and fast changing load. Therefore, the controller to be preferred for a D-STATCOM should have a nonlinear and robust structure. In this study, a neuro-fuzzy controller (NFC), which is a robust and nonlinear controller, is proposed for dq-axis current control of a D-STATCOM. A DSP-based experimental setup is built for real-time control. The basic building block of the experimental setup is a three-level cascaded inverter. This inverter is constructed by using three IPM intelligent modules. A DS1103 controller card is used for real-time control of the D-STATCOM's experimental setup. The control algorithm is prepared in MATLAB/Simulink software and loaded to the DS1103 controller card. The performance of the NFC current-controlled D-STATCOM is tested for different load conditions: no load to full inductive, no load to full capacitive, full inductive to full capacitive, and full capacitive to full inductive. For this aim, the reactive current setpoint is changed as a step. The experimental results are presented to show the efficiency of the proposed controller under different load conditions.