N-methyl-D-aspartate receptors (NMDARs) are an essential target for the analgetic action of tricyclic antidepressants (TCAs). Their therapeutic blood concentrations achieve 0.5–1.5 μM, which, however, are insufficient to cause in vitro the open-channel block known as the only effect of TCAs on NMDARs. Whereas structures of amitriptyline (ATL), desipramine (DES), and clomipramine (CLO) are rather similar these compounds manifest different therapeutic profiles and side effects. To study structure-activity relationships of DES and CLO on NMDARs, we measured IC50s as a function of extracellular calcium ([Ca2+]) and membrane voltage (Vm) of NMDAR currents recorded in cortical neurons. Here two components of TCA action on NMDARs are described, which could be characterized as the Ca2+-dependent inhibition and the open-channel block. DES demonstrated a profound Ca2+-dependent inhibition of NMDARs, while the CLO effect was weak. DES IC50 exhibited an e-fold change with a [Ca2+] shift of 0.59 mM, which is consistent with ATL. The Ca2+ dependence of NMDAR inhibition by DES disappeared in BAPTA loaded neurons, suggesting that Ca2+ acts from the inside. Since CLO differs from DES and ATL by the presence of Cl-atom in the structure, most likely, this is the atom which is responsible for the loss of pronounced [Ca2+] dependence. As for the NMDAR open-channel block, both DES and CLO were about 5-folds more potent than ATL due to their slow rates of dissociation either from open and closed states. DES demonstrated stronger Vm-dependence than CLO, suggesting a deeper location of the DES binding site within the ion pore. Because DES and CLO differ from ATL by the nitrogen-containing tricycle, presumably this moiety of the molecules determines their high-affinity binding with the NMDAR channel, while the aliphatic chain mono-methyl amino-group of DES allows a deep permeation in the channel. Thus, different structure-activity relationships of the Ca2+-dependent inhibition and Vm-dependent open-channel block of NMDARs by DES and CLO suggest that these processes are independent and most likely may represent an action on different molecular targets. The proposed model of TCA action on NMDARs predicts well the experimental values of IC50s at physiological [Ca2+] and within a wide range of Vms.