Although it has not been well-publicized, it has been possible to run an O-Plan agent remotely on the Internet since at least version 1.2. Indeed, the O-Plan planning and execution agents have always communicated in this way. In version 1.2, the link used Internet-domain sockets; version 2.0 introduced a simpler mechanism based on rsh.
[Three-agent architecture.]
However, certain limitations have only gradually been removed. The 1.2 execution agent could be used remotely, but not the planner, because the built-in task-assigner occupied the communication channels. This limitation was removed in version 2.0. In addition, all of the agents needed to display windows and interact with a user. It wasn't possible to run O-Plan invisibly in the background, and some things a program might want to access or control were available only via the windows. These problems are being addressed in the current development version.