I've had huge success with IP Webcam, especially since it appears to be a standard MJPEG stream, so very hackable (in the good sense). The web interface also crucially adds UI for focusing, in addition to the customizable in-phone UI.
Have you encountered a similar hurdle?
If you really need, run a reverse proxy only on the /video endpoint. For I don't know how hardened the focus and automation control APIs are.