I think what you want are pty's? Or, another option would be unix domain sockets.
The answer to your first question is "no"... not really. When you write out to a tty, that output is sent out to the real device and not available to be read back.
The list of tty's on a system is: lsdev -Cctty
Creating tty's at run time is possible but not really what you want either. A tty is a child of a serial port and you can not add serial ports arbitrarily. They are real things. With AIX and Power systems, you can add devices while the system is up (hot swap) but that is getting (I'm assuming) way far off your original topic.
The basic different between a pty and a unix domain socket is a pty mimics the output and input process of a real tty in one direction. This is what telnet, rlogin, ssh, and many other daemons use when connections come in. It is easy to make ksh believe that it has a real tty by using pty's. If you don't need that, then they are added trouble that you don't need. Find a link on how to create and use a Unix domain socket and you will have what you need (or a pipe but a pipe requires a parent / child relationship which I assume you do not have).