Not really hard, don't worry
An AGD suite has the following parts:
1) The "engine". An assembler source with all the core routines.
2) A "compiler". Programmed in C and responsible of reading the AGD source and traslating every command or sequence of commands in functional blocks, generating inline ASM statements of simply calling core routines. It adds to the result the engine file (obviously) and the parsed data of the graphic resources. The final result is an asm file that only needs to be compiled.
That's the main part and probably the most common points between machines. There's also a "build" script (CMD, Shell, Perl, whatever) used to invoke all the toolchain and generate the desired final distribution for the machine, maybe even to build a media and launch an emulator.
What do you have now?. Being finished the MSX port you only need to know what are you going to do with your memory map (Einstein seems to have a great amount of easy managed RAM, from 0100 to E0FF, right?) and of course how to map every piece of hardware from MSX to Einstein, VDP and PSG seems only a matter of change some EQU constants an map different ports and the "biggest" problem, the keyboard reading, completely different between machines. Sincerely I can't think of any more great differences.
The compiler will be easy to port as the amount of new code generated is small (some keyboard related routines) and there's no reason to change funtion names, so most of the glue code will remain unchanged.
I think that getting a compiled COM from an AGD would be a relatively easy task
You can ask whatever you want, of course...
