MPC, or secure multi-party computation, is a powerful class of cryptographic primitives that enables a host of privacy-preserving systems. In MPC, $n$ parties each have an input $s_i$, and aim to compute a joint function $f(s_1, …, s_n)$. Because parties can only learn the function’s output, each party’s input is entirely hidden from the rest. In recent years, milestones in the efficiency of MPC protocols have opened a new door: leveraging MPC in broad-reaching, user-facing applications. But designing and implementing an MPC protocol is only the first step to deploying an end-to-end MPC application. The authors of this post are working on real-world MPC systems, some that serve many millions of users. In this piece, we bring together our respective experiences to collectively reflect on the following question: What are the most significant challenges to the development and deployment of MPC applications?