By now, you should have either acquired an EEG or drilled some holes and closed them again. But recording mental information is not quite enough. You will need some method to amplify and filter your material, and some rules on how to do that. There are complex mathematical algorithms to extract meaningful information out of nerve cell noise in real time (more), depending on which signal you have chosen. These algorithms will also have to be able to distinguish electrical "noise", which is produced by eyes and muscles, among others.
The equipment to do all that currently is huge and expensive, so one of the biggest tasks for you will be to downsize this stuff. One day it might be possible to put what currently needs a whole table on a small chip, that communicates wirelessly with the prosthesis. Until then, you will have to live with wires coming out of your head, and some vehicle to push around. Help is already on its way, though. Prototypes of instrumentation neurochips that might even be implantable are out there (more).

The signals / build BCI 1 / build BCI 2 /  build BCI 3
components > build BCI 2
BCI scheme Graz-BCI
- have larger BCI-scheme -


Even once you have developed a small enough "instrumentation neurochip" some problems remain. It will have to perform analog and digital processing, which needs energy that has to come from somewhere.

Economical devices like the latest pacemakers for the heart can be reloaded magnetically through the skin, but your chip will need quite a lot fuel, and after all there is also the skull in between the chip the loading machine.

Furthermore it will have to have a telemetry unit to talk to your robot arm, or at least a wireless communication protocoll to talk to another computer in your pocket. All this won't make the whole thing smaller, either...

 

And you'll need some energy, won't you?

Learn about definitions of BCI, how to build your own one, and what the history of BCIs was like.