Aims. The derivation of spectroscopic parameters for M dwarf stars is very important in the fields of stellar and exoplanet characterization. The goal of this work is the creation of an automatic computational tool, able to derive quickly and reliably the T$_{\mathrm{eff}}$ and [Fe/H] of M dwarfs by using their optical spectra, that can be obtained by different spectrographs with different resolutions. Methods. ODUSSEAS (Observing Dwarfs Using Stellar Spectroscopic Energy-Absorption Shapes) is based on the measurement of the pseudo equivalent widths for more than 4000 stellar absorption lines and on the use of the machine learning Python package "scikit-learn" for predicting the stellar parameters. Results. We show that our tool is able to derive parameters accurately and with high precision, having precision errors of ~30 K for T$_{\mathrm{eff}}$ and ~0.04 dex for [Fe/H]. The results are consistent for spectra with resolutions between 48000 and 115000 and SNR above 20.