The reaction 12C(α,γ)16O is of paramount importance for the nucleosynthesis of heavier elements in stars. It takes place during helium burning and determines the abundance of 12C and 16O at the end of this burning stage and therefore influences subsequent nuclear reactions. Currently the cross section at astrophysically relevant energies is not known with satisfactory precision. Due to the low cross section of the reaction, low background, high beam intensities and target thicknesses are necessary for experiments. Therefore a new laboratory hosting a 5 MV ion accelerator, was built in the shallow-underground tunnels of Felsenkeller. The main background component in such laboratories was investigated with a muon telescope in this thesis. It was found, that the rock overburden of about 45 m vertical depth reduces the muons by a factor of about 40 compared to the surface. Furthermore the results of the measurements were compared to a simulation based on the geometry of the facility and showed good agreement. In the next step the accelerator was put into operation. Since the experiment on 12C(α,γ)16O will be done in inverse kinematics, an intense carbon beam is necessary to reach sufficient statistics. For this, the creation and extraction of carbon ions in an external ion source was improved. The external source now provides steady currents of 12C− of above 100 μA. In the following the transmission through the accelerator and the high-energy beamline was tested with a beam restricted in width. The pressure of the gas stripper in the centre of the accelerator and the parameters of different focusing elements after the accelerator were varied. It was found, that for a desired carbon beam energy of below 9 MeV, the 2+ charge state is suited best, where up to 35% of the inserted beam could be transmitted. To ease the planning of future experiments and aid the analysis of the data, the target chamber and two different kinds of cluster detectors were modelled in Geant4. The low-energy region was verified by comparing the simulations to measurements with radioactive calibration sources. Deviations for the detectors were below 10% without target chamber, and up to 30% for individual germanium crystals of the Cluster Detectors with the target chamber. A first test measurement was undertaken to investigate the capabilities of the new laboratory. Solid tantalum targets implanted with 4 He were prepared. An ERDA analysis of the used solid targets showed contaminations with carbon and oxygen. These led to beam-induced background in the region of interest during the irradiation. Then the targets were irradiated with a carbon beam at two different energies. While no clear signal of 12C(α,γ)16O could be observed, the beam could be steered on the target for the whole duration of the beam time spanning five days. Problems during this test, like low beam current, were identified. These could be partly remedied in the scope of this thesis. Suggestions for improvements for a second test run were developed as well.