Composite Calculus is a formal arithmetic system built on Laurent polynomial algebra where every operation preserves full provenance — no information is ever lost, even in traditionally destructive operations. The system models numbers as coefficient-carrying objects across a dimensional scale, enabling reversible computation and exact symbolic differentiation, integration, and multivariate calculus through simple algebraic manipulation. Includes a Python library composite_lib.py, a comprehensive test suite (175+ tests) and foundational papers. Key features: Provenance-preserving arithmetic with full reversibility Universal calculus machine via dimensional coefficient algebra Constructive infinitesimal system — fully computable, no abstract foundations required Multivariate extensions for partial derivatives and PDEs Repository: https://github.com/tmilovan/composite-machine