Wireless low-power, multi-hop networks are exposed to numerous attacks also due to their resource-constraints. While there has been a lot of work on intrusion detection systems for such networks, most of these studies have considered only a few topologies, scenarios and attacks. One of the reasons for this shortcoming is the lack of sufficient data traces that are required to train many machine learning algorithms. In contrast to other wireless networks, multi-hop networks do not contain one entity that can capture all the traffic which makes it more difficult to acquire such traces. In this paper we present Multi-Trace. Multi-Trace extends the Cooja simulator with multi-level tracing facilities that enable data logging at different levels while maintaining a global time. We discuss the opportunities that traces generated by Multi-Trace enable for researchers interested in input for their machine learning algorithms. We present experiments that show the efficiency with which Multi-Trace generates traces. We expect Multi-Trace to be a useful tool for the research community.