Hi,
For stars in the Legacy Survey, you can search the main catalog called "tractor". So if you are using the 8th data release, the table would be: ls_dr8.tractor
You can start with just the "north" (tractor_n) or "south" (tractor_s) depending on how large of an area you will need.
The full list of column descriptions is here: https://datalab.noirlab.edu/query.php?name=ls_dr8.tractor
It's a rather long list, but of interest to you, you will probably need both the g_mag (in AB magnitudes) and snr_g (signal-to-noise ratio in g-band) to make sure they are well-measured. I would also recommend to use TYPE='PSF' for point sources, and please note that there is not a perfect definition for what is a star versus not a star (such as small unresolved galaxies), so one color cut used with the Legacy Surveys could be:
z_w1 < (0.8*r_z-1.0)
Where z_w1 is the pre-computed z-W1 color from Data Lab, and r_z is the r-z color. It will be interesting for you to check with the image cutouts that you are indeed finding objects that look like stars!
Other recommendations for the data quality are to use brick_primary=1 and maskbits=0 to avoid data problems that get a flag encoded in the maskbits.
Lastly, you may want to start with a "LIMIT 100" statement or some manageable number until you are fully happy with your query and the columns that it returns.