The Chebyshev rational approximation method (CRAM) computes the matrix exponential by applying the best rational approximation (of order k) to ex on (−∞,0] directly to the matrix. Published CRAM coefficients are available up to order 48. Applications such as coupled depletion-transport calculations, extended-precision arithmetic, nuclear safeguards verification, or nuclear arms control and disarmament verification may require higher-order coefficients or higher precision, yet the code to compute them has never been published. Pusa described the underlying Remez algorithm in detail but did not release an implementation. I present an open-source C++ implementation that solves the nonlinear equioscillation system in a Möbius-transformed variable, enabling the efficient determination of CRAM coefficients of arbitrary even order and precision. For orders above 48, where the direct algorithm fails due to Newton degeneracy, I introduce an order-continuation scheme that chains converged solutions from order k to order k+2 using equioscillation-derived alternation points. Using this approach, I reproduce all published coefficients and extend the computation to order 100. The method is demonstrated on the 232Th decay chain spanning 24 orders of magnitude in eigenvalue stiffness. A second example, a ten-nuclide actinide burnup chain, shows a trace 235U concentration that CRAM-48 cannot resolve but CRAM-100 recovers. The source code and all computed coefficients are publicly available.