Background: Monitoring central hemodynamics is essential in critically ill patients and less invasive techniques are needed. In this study, the clinical and technical performance of a new non-invasive cardiac output monitor (NICO) based on partial CO2 rebreathing technique and a modified Fick equation were evaluated. The various sources of possible errors in measurement of cardiac output (CO), carbon dioxide production (V˙CO2) and pulmonary shunt were also assessed. Methods: Simultaneous measurements of CO with partial CO2 rebreathing technique (COnico) and thermodilution (COtd) were performed in 15 patients during major surgery or in the ICU. Pulmonary shunt was estimated from this device and compared to values obtained by standard shunt formula. The accuracy of V˙CO2 measurements was assessed in a mechanical lung model. Results: A good correlation was found between COnico and COtd (r = 0.96, within-subject correlation r = 0.88) with a small underestimation of cardiac output by the NICO of 0.04 L/min, limits of agreement (± 2 SD) being − 1.68 and 1.76 L/min. In hemodynamic unstable patients the method closely tracked changes in CO. Pulmonary shunt was underestimated by approximately 11%-units compared to standard shunt calculations using arterial and mixed venous blood gases. We also observed an underestimation in V˙CO2 measurements. Conclusion: Clinical evaluation shows that partial CO2 rebreathing technique provides a useful and accurate non-invasive estimate of cardiac output. Although this technique cannot fully replace the pulmonary artery catheter, it may be used to monitor central hemodynamics in a large number of critically ill patients.