Recent developments in enhanced sampling methods showed that it is possible to reconstruct ligand unbinding pathways with spatial and temporal resolution inaccessible to experiments. Ideally, such techniques should provide an atomistic definition of possibly many reaction pathways, because crude estimates may lead either to overestimating energy barriers, or inability to sample hidden energy barriers that are not captured by reaction pathway estimates. Here we provide an implementation of a new method [Rydzewski and Valsson, J. Chem. Phys. 150, 221101 (2019)] dedicated entirely to sampling the reaction pathways of the ligand–protein dissociation process. The program, called maze, is implemented as an official module for PLUMED 2, an open source library for enhanced sampling in molecular systems, and comprises algorithms to find multiple heterogeneous reaction pathways of ligand unbinding from proteins during atomistic simulations. The maze module requires only a crystallographic structure to start a simulation, and does not depend on many ad hoc parameters. The program is based on enhanced sampling and non-convex optimization methods. To present its applicability and flexibility, we provide several examples of ligand unbinding pathways along transient protein tunnels reconstructed by maze in a model ligand–protein system, and discuss the details of the implementation. Program SummaryProgram title:mazeProgram files doi:http://dx.doi.org/10.17632/x5zsgzxcnx.1Licensing provisions: LGPL-3.0Programming language: C++11Nature of problem: Finding heterogeneous unbinding reaction pathways of ligand transport processes such as dissociation and diffusion in ligand–protein systems during molecular dynamics simulations.Solution method: Enhanced sampling molecular dynamics method with non-convex optimization techniques performed on-the-fly during biased simulations.Unusual features: maze is a module for the PLUMED 2 software, it must be used in conjunction with this software, embedded in a molecular dynamics code such as Gromacs, NAMD, LAMMPS, OpenMM, and Amber.Additional comments: Program Github repository https://maze-code.github.io