Ecosystems of multiple blockchains are now a reality. Multi-chain applications and protocols are perceived as necessary to enable scalability, privacy, and composability. Despite being a promising emerging area, we have been witnessing devastating attacks on cross-chain bridges that have caused billions of dollars in losses, and no apparent solution seems to emerge from the ongoing chaos. In this paper, we present our contribution to minimizing bridge attacks, by monitoring a cross-chain model. In particular, we aggregate cross-chain events into cross-chain transactions, and verify if they follow a set of cross-chain rules, which then generate a model. We propose Hephaestus, the first cross-chain model generator that captures the operational complexity of cross-chain applications. Hephaestus can generate cross-chain models from local transactions in different ledgers, realizing arbitrary cross-chain use cases and allowing operators to monitor their applications. Monitoring helps identify outliers and malicious behavior, which can enable programmatically stopping attacks (“a circuit breaker”), including bridge hacks. We conduct a detailed evaluation of our system, where we implement a cross-chain bridge use case. Our experimental results show that Hephaestus can process 600 cross-chain transactions in less than 5.5 seconds in an environment with two blockchains using sublinear storage, paving the way for more resilient bridge designs.