The file coil_def.dat contains the geometry of various types of MEG sensors, independently from any acquisition device.
The functions mne_add_coil_defs.m or ctf_add_coil_defs.m apply this geometry to the description of an acquisition device (position and orientation of each sensor in the MEG system), as it can be found in a .fif file (Elekta) or .res4 file (CTF).
https://github.com/mne-tools/mne-matlab/blob/master/matlab/mne_add_coil_defs.m
https://github.com/brainstorm-tools/brainstorm3/blob/master/toolbox/io/ctf_add_coil_defs.m
To get started, you need the 3D positions and orientations of your sensors relatively to the head of your subject, Braintorm or MNE-Python do not provide this information. Then you can use Brainstorm or MNE-Python functions to compute all the integration points (ie. the info you can find in the Brainstorm channel.mat file) needed for computing your forward model.
To get the positions and orientations of the MEG sensors relatively to the head of the subject, most MEG systems use head localization coils (3 for CTF and 4D, 4 for Elekta, 5 for Yokogawa, etc). I think you can find online documentation about this technique easily.