The main focus of the CBM experiment (FAIR, Germany) is the measurement of very rare probes at interaction rates of up to 10 MHz. The experiment will operate with a data flow of up to 1 TB/s and requires full online event reconstruction and selection. This is a task of the First Level Event Selection (FLES). A standalone FLES package has been developed for the CBM experiment. It contains modules for all reconstruction stages: track finding, track fitting, short-lived particle finding and selection. Reconstruction of about 50 types of particle decay channels is currently implemented. The algorithms are local with respect to data and their implementation is both vectorized (SIMD) and parallelized between CPU cores. For the track reconstruction the Cellular Automaton (CA) and the Kalman filter (KF) algorithms are used, that allows to achieve a high track reconstruction efficiency of up to 97% and a track parameters quality with 1% momentum resolution. The KF particle finder has a high efficiency with an optimal signal to background ratio. The FLES package shows a strong scalability on many-core systems and a processing speed of 1700 events per second on an Intel based computer with 80 cores. The investigation was done based on simulated minimum bias Au-Au UrQMD collisions at 25 AGeV with a realistic detector response.