Command line usage is normally appropriate when you wish to handle a whole process as one, and record all its interactions to the same file. This is generally only the case for command-line and client-server interception, not for intercepting Python code, where you want different interaction for different tests within the same process.
The basic plan is to run the "capturemock" program. Which arguments to give it can be discovered by running it alone or with --help:
$ capturemock --help
Usage: capturemock [options] ...
CaptureMock command line program. Records and replays interaction
defined by stuff in its rc file
-h, --help show this help message and exit
-m MODE, --mode=MODE CaptureMock mode. 0=replay, 1=record, 2=replay if
possible, else record
-p FILE, --replay=FILE
replay traffic recorded in FILE.
-f DIR, --replay-file-edits=DIR
restore edited files referred to in replayed file from
-r FILE, --record=FILE
record traffic to FILE.
-F DIR, --record-file-edits=DIR
store edited files under DIR.
-R RCFILES, --rcfiles=RCFILES
Read configuration from given rc files, defaults to
Basic usage consists normally of
$ capturemock --record some_file.mock my_program -a -b -c
$ capturemock --replay some_file.mock my_program -a -b -c