The CNAP system (CNSystems Medizintechnik AG, Graz, Austria) provides noninvasive continuous arterial pressure measurements by using the volume clamp method. Recently, an algorithm for the determination of cardiac output by pulse contour analysis of the arterial waveform recorded with the CNAP system became available. We evaluated the agreement of the continuous noninvasive cardiac output (CNCO) measurements by CNAP in comparison with cardiac output measurements invasively obtained using transpulmonary thermodilution (TDCO). In this proof-of-concept analysis we studied 38 intensive care unit patients from a previously set up database containing CNAP-derived arterial pressure data and TDCO values obtained with the PiCCO system (Pulsion Medical Systems SE, Feldkirchen, Germany). We applied the new CNCO algorithm retrospectively to the arterial pressure waveforms recorded with CNAP and compared CNCO with the corresponding TDCO values (criterion standard). Analyses were performed separately for (1) CNCO calibrated to the first TDCO (CNCO-cal) and (2) CNCO autocalibrated to biometric patient data (CNCO-auto). We did not perform an analysis of trending capabilities because the patients were hemodynamically stable. The median age and APACHE II score of the 22 male and 16 female patients was 63 years and 18 points, respectively. 18 % were mechanically ventilated and in 29 % vasopressors were administered. Mean ± standard deviation for CNCO-cal, CNCO-auto, and TDCO was 8.1 ± 2.7, 6.4 ± 1.9, and 7.8 ± 2.4 L/min, respectively. For CNCO-cal versus TDCO, Bland-Altman analysis demonstrated a mean difference of +0.2 L/min (standard deviation 1.0 L/min; 95 % limits of agreement -1.7 to +2.2 L/min, percentage error 25 %). For CNCO-auto versus TDCO, the mean difference was -1.4 L/min (standard deviation 1.8 L/min; 95 % limits of agreement -4.9 to +2.1 L/min, percentage error 45 %). This pilot analysis shows that CNCO determination is feasible in critically ill patients. A percentage error of 25 % indicates acceptable agreement between CNCO-cal and TDCO. The mean difference, the standard deviation, and the percentage error between CNCO-auto and TDCO were higher than between CNCO-cal and TDCO. A hyperdynamic cardiocirculatory state in a substantial number of patients and the hemodynamic stability making trending analysis impossible are main limitations of our study.