The risk or value process of an insurance company, modelled by a Cramer-Lundberg model, is supposed to be controlled by a reinsurance share, that is a part of the risk is undertaken, but also premium has to be divided. The aim is to control this reinsurance level in way, that the discounted value of the risk process maximizes. First, the process is approximated by a diffusion process, then stochastic control theory is used to find an optimal value function and an optimal control. Non-cheap reinsurance and a bankruptcy value are also considered. In the last part of the thesis Monte-Carlo simulation is used to calculate examples and verify the solution.