Monte Carlo (MC) calculations are a fundamental tool for the investigation of ionization chambers (ICs) in radiation fields, and for calculations in the scope of IC reference dosimetry. Geant4, as used for the toolkit TOPAS, is a major general purpose code, generally suitable for investigating ICs in primary proton beams. To provide reliable results, the impact of parameter settings and the limitations of the underlying condensed history (CH) algorithm need to be known. A Fano cavity test was implemented in Geant4 (10.03.p1) for protons, based on the existing version for electrons distributed with the Geant4 release. This self-consistent test allows the calculation to be compared with the expected result for the typical IC-like geometry of an air-filled cavity surrounded by a higher density material. Various user-selectable parameters of the CH implementation in the EMStandardOpt4 physics-list were tested for incident proton energies between 30 and 250 MeV. Using TOPAS (3.1.p1) the influence of production cuts was investigated for bare air-cavities in water, irradiated by primary protons. Detailed IC geometries for an NACP-02 plane-parallel chamber and an NE2571 Farmer-chamber were created. The overall factor f Q as a ratio between the dose-to-water and dose to the sensitive air-volume was calculated for incident proton energies between 70 and 250 MeV. The Fano test demonstrated the EMStandardOpt4 physics-list with the WentzelIV multiple scattering model as appropriate for IC calculations. If protons start perpendicular to the air cavity, no further step-size limitations are required to pass the test within 0.1%. For an isotropic source, limitations of the maximum step length within the air cavity and its surrounding as well as a limitation of the maximum fractional energy loss per step were required to pass within 0.2%. A production cut of ⩽5 μm or ∼15 keV for all particles yielded a constant result for f Q of bare air-filled cavities. The overall factor f Q for the detailed NACP-02 and NE2571 chamber models calculated with TOPAS agreed with the values of Gomà et al (2016 Phys. Med. Biol. 61 2389) within statistical uncertainties (1σ) of <0.3% for almost all energies with a maximum deviation of 0.6% at 250 MeV for the NE2571. The selection of hadronic scattering models (QGSP_BIC versus QGSP_BERT) in TOPAS impacted the results at the highest energies by 0.3% ± 0.1%. Based on the Fano cavity test, the Geant4/TOPAS Monte Carlo code, in its investigated version, can provide reliable results for IC calculations. Agreement with the detailed IC models and the published values of Gomà et al can be achieved when production cuts are reduced from the TOPAS default values. The calculations confirm the reported agreement of Gomà et al for [Formula: see text] with IAEA-TRS398 values within the given uncertainties. An additional uncertainty for the MC-calculated [Formula: see text] of ∼0.3% by hadronic interaction models should be considered.