The Survey Science Centre of the XMM-Newton satellite released the first incremental version of the 2XMM catalogue in August 2008. Containing more than 220 000 X-ray sources, the 2XMMi was at that time the largest catalogue of X-ray sources ever published and thus constitutes an unprecedented resource for studying the high-energy properties of various classes of X-ray emitters such as AGN and stars. Thanks to the high throughput of the EPIC cameras on board XMM-Newton accurate positions, fluxes, and hardness ratios are available for a substantial fraction of the X-ray detections. The advent of the 7th release of the Sloan Digital Sky Survey offers the opportunity to cross-match two major surveys and extend the spectral energy distribution of many 2XMMi sources towards the optical bands. This implies building extensive homogeneous samples with a statistically controlled rate of spurious matches and completeness. We here present a cross-matching algorithm based on the classical likelihood ratio estimator. The method developed has the advantage of providing true probabilities of identifications without resorting to heavy Monte-Carlo simulations. Over 30,000 2XMMi sources have SDSS counterparts with individual probabilities of identification higher than 90%. At this threshold, the sample has only 2% spurious matches and contains 77% of all expected SDSS identifications. Using spectroscopic identifications from the SDSS DR7 catalogue supplemented by extraction from other catalogues, we build an identified sample from which the way the various classes of X-ray emitters gather in the multi dimensional parameter space can be analysed and later used to design a source classification scheme. We illustrate the interest of this clean source sample by investigating two scientific use cases. In the first example we show how these multi-wavelength data can be used to search for new QSO2s. Although no specific range of observed properties allows us to efficiently identify Compton Thick QSO2s, we show that the prospects are much better for Compton Thin AGN2 and discuss several possible multi-parameter selection strategies. In a second example, we confirm the hardening of the mean X-ray spectrum with increasing X-ray luminosity on a sample of over 500 X-ray active stars and reveal that on average X-ray active M stars display bluer g − r colour indexes than less active ones. Although this catalogue of 2XMM-SDSS sources cannot be used directly for statistical studies, it nevertheless represents an excellent starting point to select well defined samples of X-ray-emitting objects.