Cooperation among CBR agents involves exploiting the set of precedents in the collective memories of all the agents for use in similarity-based reasoning. There are two general ways to do so: DistCBR and ColCBR. Intuitively, both DistCBR and ColCBR are based on solving a problem by reusing the knowledge learned by other CBR agents. Given an agent (the originator) trying to solve a problem, the difference between both modes is in the similarity-based reasoning method used: that of the originator or that of the CBR agent that is helping the originator.
transmitting the problem and the task
to be achieved to another agent 
. Agent uses its own CBR method
and its case base 
to
achieve the task and send the results back
to agent . In case of
failure, a failure token is sent back and
can iterate the
cooperation tasks with the next agent of its preference.
transmitting the method that
is to be used to solve a task, in addition to the problem and the task
to be achieved, to another agent . Agent will use its case
base and the method
sent by to achieve the
task and send
back the results. In other words, the originator is using the memory
of the other agents as an extension of its own - as a collective
memory - by means of being able to impose on other agents the use of
its own CBR methods. In case of failure, a failure token is sent back
and can iterate the
cooperation tasks with the next agent of its
preference.
From the standpoint of implementing these cooperation modes, we can say that DistCBR is supported by the remote evaluation capability and ColCBR is supported by remote programming (or mobile code) capability of Plural Noos.