In the past, several Soar projects have been interested in the use of
multiple agents.  Up to this point, researchers have developed
separate systems by modifying their own copy of the existing Soar
source.  While helpful for the task at hand, these modifications were
not added to the mainstream Soar development.  With the impetus of the
Darpa-funded IFOR effort, a multi-agent version of Soar 6 has been
created.
 
This version of Soar 6 enables the user to create an arbitrary number
of agents, each of which has a separate memory.  All the agents run
together within a single process to minimize the overhead of
communicating among the agents of the system.  [Consequently, this
design does not add any additional operating system requirements on
the existing Soar 6 code.]
 
If the user does not desire the use of multiple agents, then the use
of Soar 6 will be the same as before (except for I/O code, see the
next section).

