bombdiggity is on it. Record as you would normally. Adjust the relative levels of the two sources, and the overall summed level afterwards.
If you are just getting going with this, and have not already developed a large time/experience/(and often financial) commitment to the workflow of a particular piece of software, I'd go with Sloan's suggestion of using Reaper. It's inexpensive, very flexible, is in steady development, and has a large user community.
My advice on how to go about mixing the two sources regardless of what program is used is this-
First listen to each source independently and do whatever needs to be done on each individually: panoramic balance, eq, dynamics, fixes, whatever.. anything expect processes which change the run-length of one source verses the other.
I assume both sources were recorded using one recorder and are already in sync with each other, so you won't need to shrink/stretch one or otherwise sync them. If not you'll need to do that next. Even if both sources were recorded to one machine, sometimes its to make time-shift delay adjustments between the two sources. Now is the time to do that.
At that point, you can play around with finding an optimal level balance between the two sources for the sum. It's rarely going to be exactly the same level for both of them.
Then make any collective adjustment tweaks necessary to the combination of the two, including their overall summed level. The amount by which the signal level will increase when the two sources are summed is not going to be simply predictable beforehand. It depends on how precisely related the information is in the channels which are being summed.