Context. Searching for planets in open clusters allows us to study the effects of dynamical environment on planet formation and evolution. Aims. Considering the strong dependence of planet frequency on stellar metallicity, we studied the metal rich old open cluster NGC 6791 and searched for close-in planets using the transit technique. Methods. A ten-night observational campaign was performed using the Canada-France-Hawaii Telescope (3.6m), the San Pedro Mártir telescope (2.1m), and the Loiano telescope (1.5m). To increase the transit detection probability we also made use of the Bruntt et al. (2003) eight-nights observational campaign. Adequate photometric precision for the detection of planetary transits was achieved. Results. Should the frequency and properties of close-in planets in NGC 6791 be similar to those orbiting field stars of similar metallicity, then detailed simulations foresee the presence of 2-3 transiting planets. Instead, we do not confirm the transit candidates proposed by Bruntt et al. (2003). The probability that the null detection is simply due to chance coincidence is estimated to be 3%-10%, depending on the metallicity assumed for the cluster. Conclusions. Possible explanations of the null-detection of transits include: (i) a lower frequency of close-in planets in star clusters; (ii) a smaller planetary radius for planets orbiting super metal rich stars; or (iii) limitations in the basic assumptions. More extensive photometry with 3-4m class telescopes is required to allow conclusive inferences about the frequency of planets in NGC 6791. Comment: 23 pages, 23 figures, A&A accepted