Analysis of the heart rate variability (HRV) gains more and more importance in the assessment of training practice and welfare in equine industry. It relies on mathematical analyses of reliably and accurately measured variations in successive inter-beat intervals, measured as RR intervals. Nowadays, the RR intervals can be obtained through two different techniques: a heart rate meter (HRM) or an electrocardiogram (ECG). The agreement and reliability of these devices has not been fully assessed, especially for recordings during exercise. The purpose of this study was to assess the agreement of two commercially available devices using the two mentioned techniques (HRM vs ECG) for HRV analysis during a standardized exercise test. Simultaneous recordings obtained during light exercise and during canter with both devices were available for 36 horses. Data were compared using a Bland–Altman analysis and the Lin’s coefficient. The agreement between the assessed HRV measures from the data obtained from the ECG and HRM was acceptable only for the mean RR interval and the mean heart rate. For the other studied measures (SDNN, root mean square of successive differences, SD1, SD2, low frequency, high frequency), the agreement between the devices was too poor for them to be considered as interchangeable in these recording conditions. The agreement tended also to be worse when speed of the exercise increased. Therefore, it is necessary to be careful when interpreting and comparing results of HRV analysis during exercise, as the results will depend upon recording devices. Furthermore, corrections and data processing included in the software of the devices affect largely the output used in the subsequent HRV analysis; this must be considered in the choice of the device.