Below is my understanding of the situation. I have never done it myself but have given it a little thought and research. When the dual channel NetSDR (RFSpace) was announced I thought this might be a cool direction to go. While it is a heck of a nice thought, at a really simple level the answer to your question is no.
First is the two antenna / sample channel issue. With only two inputs / antennas and if everything else is right you end up with two possible directions for any signal, it does not narrow down to one direction. Grating lobes are an issue, so even those 2 directions can be in question and the array probably has general directions of optimal and suboptimal performance. To get one direction you need at least 3 inputs / antennas, and 4 is more or less the best staring point. 2 antennas would be great for steering a lobe or a null though. Spacing of the antennas are going to realistically limit bands of operation. Feedline variables must be known but should be able to be calibrated out.
Then comes questions about the inputs of the SDR. Is it separate ADC’s? Or is it one device? If it is multiple ADCs are they disciplined to a single source so that the phase relationship (time of arrival) of the inputs are known? If each ADC is free running then the phase relationship of the signals in each is in question, and direction cannot be derived.
Then is the aspect of the software at the hobby level, at the pro level there are lots of off the shelf options. At the hobby level there are several people playing around with this, and GNU Radio has some interesting stuff, but there is no simple interface you can use right now to display a direction, even if you have everything else right. Everyone who is doing this appears to be writing their own stuff, possibly building off others works as a starting point.
T!