The master-agent communication now is not stable, we need to run the agent first then master to get the communication working. This should be made flexible when we can run the master first then trigger agent running.