A wealth of data collected over the past three decades has demonstrated that volcanic unrest is often associated with elevated levels of seismicity. Volcano seismic networks commonly record intense swarms of earthquakes in the weeks to months before eruptions; peak rates of more than one event per minute are common. The ability to readily detect and classify these signals is crucial to effective monitoring operations and hazard assessment. The sheer volume of information collected, however, poses a challenge to volcano observatories because of the unrealistically large number of staffs required for manual inspection of these data. Here, we present Recursive Entropy Method of Segmentation (REMOS), a computationally efficient Python workflow used to detect, extract, and classify volcanic earthquakes starting from raw continuous waveform data. Within REMOS, seismograms are first analyzed using the well‐established short‐term average/long‐term average method to identify trigger times of candidate earthquakes. A new algorithm based on measurements of seismic energy and minimum entropy is then used to investigate large amounts of earthquake triggers and to discriminate and parse events into individual waveforms for further analyses. REMOS also includes a facility for classification of the extracted waveforms based on simple frequency‐domain metrics. Finally, the results can be visualized using t‐distributed stochastic neighbor embedding, a technique for dimensionality reduction that is particularly well suited to inspection of high‐dimensional datasets. In this work, we demonstrate the use of REMOS with seismic data recorded in 2007 during a period of unrest and eruption at Bezyminany Volcano. Our results show that REMOS can efficiently detect, segment, and classify earthquakes at scale and at very low computational cost.