Counting absolute numbers of proteins is of major importance in biology. This would help for example to resolve the stoechiometry of complexes, to observe aggregation phenomena or to describe the minimal signal required for a specific cellular response. The only existing tools available to biologists are bulk experiments, crystallography or photobleaching steps. Each of these methods comes with its own drawbacks: by comparing global mean values, bulk experiments do not take into account quenching or differences of brightness in a population of fluorophores; crystallography only studies the protein as a crystal and hence in a totally artificial environment; finally, the detection of photobleaching steps requires a low density of relevant complexes and a low stoechiometry (up to five monomers).Based on a TIRF (total-internal reflexion fluorescence) single-molecule microscopy approach, we set out to optimise and characterise an original method to precisely count absolute numbers of proteins in live cells.We first characterise the photo-physics (contrast, blinking, dark state, pre-activation, reversible activation, thermal or pH dependant activation, bleaching, etc.) of different photo-activable dyes. Although a blinking fluorophore allows multiple localization in PALM (photo-activable localization microscopy) or STORM (stochastic optical reconstruction microscopy) imaging, it should be only counted once when counting is at stake. We also develop a specific algorithm to avoid over- or under-counting single emitters. using a herpes simplex virus which capsid is tagged with exactly 900 photo-activable dyes, we then characterised the precision of our technique. Finally, we show the applicability of our approach to current questions in biology by investigating the oligomerisation state of various cell surface receptors in live T cells.