We are studying the asymptotic behavior of multi-agent or many-particle systems as the number of agents increases. In a major difference with classical mean-field limits however, agents are not identical and in particular, which pair of agents interact and how much is prescribed by a possibly time-dependent graph of interactions. This type of models are used in a wide range of applications from coupled oscillators to networks of biological neurons. Those applications require a large number of agents or particles, making the mean-field limit attractive. But the usual notion of propagation of chaos cannot hold anymore which forces major changes in the classical theory. This is a joint work with D. Poyato and J. Soler.