This study investigates the consensus tracking problem for a class of networked non-linear multi-agent systems (NNMASs) using cloud computing. To achieve the stability and output consensus of the NNMAS and to actively compensate for two-channel network delays, a data-based cloud predictive control scheme is proposed. The design of the proposed novel control scheme, which only depends on the historical input and output data of the agents without using the explicit or implicit information of its structure, is detailed. Sufficient conditions are derived to guarantee the stability and output consensus of the closed-loop NNMAS. Both numerical simulations and cloud-based practical experiments are conducted to demonstrate the effectiveness of the proposed scheme. The outcome promotes the engineering application of cloud computing in multi-agent systems.