In the shots over his shoulder you see him using the original mouse with his right hand, and a "chord keyboard" with his left. He calls it a "keyset". That's a set of buttons, basically function keys, except they can be "played" in combinations. The chord keyboard was also an I/O device on the Xerox Alto, but AFAIK wasn't used again.

I notice also, things didn't go as expected; he makes quite a few mistakes and very charmingly apologizes. But the fact that you can make mistakes and smoothly recover from them, was an important part of the demo.

For context, by 1968 the concept of interacting directly with a computer in what we today would call command-line mode had been around for several years. The DEC PDP-1 was released in 1959; the IBM 1620 in 1960; Multics had started in 1964; the IBM 360 series began deliveries in 1965; all allowed an operator who had the privilege of being in the "machine room" to enter commands and get responses at a typewriter. The PDP-1 was alone (I think?) in having a 2-dimensional vector-based screen, but it was used for graphics (including the original Space War), not for text.

Overwhelmingly, the paradigm for working with a computer was to prepare a "job" consisting of commands and blocks of data, all in punch cards, which was run as a "batch", producing output to print a/o disk or tape. If you made a mistake in setting up your job, you found out about it later, and could only "edit" by re-punching some cards, and re-running it.

So when Englebart showed smooth interaction with data, editing and changing data ad-hoc, and making mistakes and correcting them ad-hoc without drama, on a two-dimensional screen -- that alone was a revolution. Hyper-linked media, hierarchical lists, that was icing.