After having recorded your mental signals and having squeezed them through nasty algorithms, after having solved the problems of multiple electrode implantation, power supply, and large scale information processing within a device that should not exceed a thumbnail in size, you now have reached a stage where there are merely problems of imagination...
You could start the easy way: Most current BCIs work with computer screens and a cursor or letters, the mental typewriter thing, you name it. Too easy. You would rather go for an egoshooter or a flight simulator to be played by thought. After all, if you can move a cursor, you should be able to move a plane, too (more). The next step would be interaction with the real world. Grabbing, switching machines on and off, or drive a wheelchair, those things should be achievable. Then come the prostheses and the robot arms. That is most definitely another league, though, which you should not even think of before having realised the egoshooter. That is because you will probably need a feedback loop for that (more).

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

Easy? A question of feedback, of course...

What all these "easy" applications have in common is that they rely on the eyes to provide the feedback to the brain. That is especially obvious with thought controlled screen applications: On a mental typewriter, you see the letters or the cursors, and your eyes tell the brain if it is doing well. In principle, one could you use other senses, as well. A blind BCI-user might work with certain sounds that signal on or off, right or wrong, A or B and so on.

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