Enzymes catalyze chemical reactions. The protein lowers the overall activation energy needed for the reaction compared to the same reaction in solution. There are various ‘strategies’ of enzymes to achieve this, ranging from providing a suitable scaffolding for the reaction, to the active (covalent) participation of amino acid residues in the mechanism. A detailed understanding of the reaction mechanism in the enzyme environment is often critical for the design of efficient inhibitors or the engineering of proteins to perform different tasks. The mechanisms can be studied computationally by determining the possible intermediates and how they are connected. Since the mechanisms involve bond breaking and forming, the computational study requires Quantum Mechanical (QM) methods. However, the computational cost of these methods scales unfavorably with the size of the system, and even state-of-the-art techniques cannot be applied to an entire enzyme. To reduce the computational cost of conventional QM methods, one can employ hybrid QM/MM schemes. The molecular system is partitioned into a QM region and an (inexpensive) Molecular Mechanics (MM) region. The bond breaking and formation must be localized in the QM region, while the effect of the (protein) environment is treated at the MM level of theory. Insights in the enzymatic mechanisms can be obtained from mapping the QM/MM potential energy surface. We present methods for the investigation of QM/MM potential surfaces. Besides the location of equilibrium structures, our methods also allow the location of the connecting transition states that determine the activation energies. Normal mode analysis can then be used to characterize the resulting critical points, while reaction paths can be calculated to confirm the relationships between the critical points. The strengths of our methods are in the very high level QM methods that can be incorporated in the QM/MM scheme, the rigorous description of the QM/MM potential surface, and the ability to fully optimize the critical points. We demonstrate our methods using various examples, such as the hydrogen peroxide reduction by Selenoprotein Glutathione Peroxidase and thermal isomerization of retinal in rhodopsins.