This is about the executable format, not Microsoft's Component Object Model
COM files are extremely simple executable files. They are useful for loading programs without complications like relocating symbols, reading file headers, etc. However, their simplicity can make their usefulness limited, so you may want to support something more complex but useful, like ELF, once your OS is advanced.
COM files were used by MS-DOS. They are raw binaries, meaning there is no header data. Also, COM files are restricted to the size of one segment (a real-address mode segment, 64kb), minus 256 bytes. When a COM file is loaded, it is assumed that all code and data fits in one segment, and it is loaded to offset 0x100.
MS-DOS creates and stores an info structure starting at offset 0 and ending just before 0x100, called the Program Segment Prefix (PSP). The PSP is made just before starting the COM program. More info about the PSP can be found here. PSP is only required for running DOS programs.
COM files usually terminate with a RET instruction. On DOS, the RET pops a empty word from the stack, which results in a interrupt 0x20 (DOS API Exit) being called.
To load a COM file you need to load it to a low memory location, point ds and es to the start of the file with an 0x100 org, and jump to the start, with offset 0x100.