A key question in neuroscience is how neurons compute their specialized output from their synaptic inputs. Another key question is whether this computation is shaped by non-linear dendritic integration. To answer these questions, it would be ideal to map synaptic activity while recording dendritic and somatic responses. We developed a strategy to simultaneously measure glutamate synaptic release and postsynaptic responses in L2/3 pyramidal neurons in the mouse primary visual cortex. We sparsely coexpressed two sensors: the green membrane sensor iGluSnFR3, reporting presynaptic release of glutamate, and the red intracellular calcium sensor jrGeco1a, reporting dendritic activations and somatic spiking. We imaged longitudinally from the same neurons, targeting different dendritic branches over days, and analyzed visual responses to drifting gratings. Our approach enabled mapping of visual properties of glutamatergic inputs across a large fraction of the dendritic tree. We imaged 3-11 dendritic branches per neuron (an average of 6), each bearing numerous putative glutamatergic synapses. Individual synapses were sharply tuned, yet neurons received a broadly tuned aggregate synaptic input, with individual dendrites biased to different orientations and directions. Comparing these inputs with somatic spiking responses and postsynaptic dendritic activations revealed pronounced nonlinear input-output transformations. Our findings demonstrate the feasibility of simultaneously investigating presynaptic release and postsynaptic responses in individual dendrites at the resolution of single synapses. This method allows us to explore dendritic specialization, the functional organization of synapses, and the non-linear transformations that underpin the sensory tuning of visual neurons.