Bi-directional interfacing electronics with living electrogenic cells has been used widely in neuroscience, mainly as research tool for studying brain mechanisms, the development of neuroprostheses, and medical therapies for several neurological disorders. Challenges for this interfacing include minimizing the interference of extracellular stimulation on the recorded neural signals, increasing the stimulation selectivity, integrating recording and/or stimulation capabilities in closed-loop systems, and fast online data analysis. The main objective of this thesis was to design and implement a system that enables the simultaneous stimulation and recording neural activity in vivo from multiple electrode sites, and that allows ultimately closed-loop operation, namely neural modulation based on activity data. The realized system should be miniaturized (hence microsystem) and be suitable for in vivo rodent experiments.Two different stimulation-recording systems have been implemented in the thesis. The first single-channel prototype explores the stimulation effects on recording capabilities in the case of planar silicon neural probes. We have been able to define the voltage thresholds for eliciting neural responses experimentally. Additionally, some configurations have revealed minimal interference with stimulation artifacts, allowing the observation of evoked action potentials right after the stimulation period.A novel concept of mixed-signal architecture for removing stimulus-induced electrical artifacts and allowing recordings during the stimulation phase, was further integrated in the prototype. The template subtraction algorithm has been applied in an analog-digital context, i.e. generating the template digitally after the training phase, and subtracting it from the recording signals in the stimulation phase. The topology has successfully been validated in simulations, in in vitro measurements, and, with promising results, in vivo. This system advances the possibilities for exploring brain mechanisms, particularly under multi-electrode stimulations.Optical stimulation, in which (visible) light pulses are delivered to brain tissue to elicit or inhibit neural activity, has recently opened new perspectives in brain research, allowing much higher spatial resolution and selective stimulation. In a second prototype we realized a hybrid neural interface combining optical stimulation with electrical recording capabilities in a closed-loop fashion with fast spike sorting. This system allows acquiring and classifying the data, and subsequently triggering the light pulses online. The fast operation and high data-rate signal processing are essential in improving the precision, selectivity, and effectiveness of stimulation.