This paper presents the first real time implementation of the Fitting Oscillations & One Over F (fooof) algorithm for advanced spectral analysis of magneto- and electroencephalography (M/EEG) signals in Brain-Computer Interfaces (BCIs). Compared to basic Fast Fourier Transform (FFT) methods, the fooof model offers unique advantages for M/EEG analysis: It can distinguish neurophysiologically relevant oscillatory components from (irrelevant) aperiodic 1/f-like components and also enables the tracking of changes in peak frequency, thereby providing richer and more dynamic insights into true neural activity. We test our implemented algorithm with streamed MEG and EEG simulated data with different activation levels and streamed recorded data. The algorithm shows rapid processing with an average delay of 23.30 ms (SD=2.89 ms) adding about 16.3 ms on top of the FFT calculation while using a 1 second window kernel. The method successfully detects M/EEG oscillatory activity, accommodates variations in power levels, and exhibits a high accuracy (94.3%) in streamed recorded data. However, the balance between accuracy and delay must be carefully managed, necessitating optimization of window size depending on targeted spectral bands and application use cases, which is discussed in the paper. Our real time fooof algorithm implementation, opens a promising avenue for BCI research and application given its ability to identify true, neurophysiologically relevant, neural oscillations. The algorithm may establish itself as a suitable approach for real time M/EEG BCI neuroprosthetic or neurotherapeutic applications.