Aims. The XMM-Newton distant cluster project (XDCP) aims at the identification of a well defined sample of X-ray selected clusters of galaxies at redshifts z ≥ 0.8. As part of this project, we analyse the deep XMM-Newton exposure covering one of the CFHTLS deep fields to quantify the cluster content. We validate the optical follow-up strategy as well as the X-ray selection function. Methods. We searched for extended X-ray sources in archival XMM-Newton EPIC observations. Multi-band optical imaging was performed to select high redshift cluster candidates among the extended X-ray sources. Here we present a catalogue of the extended sources in one the deepest ∼250 ks XMM-Newton fields targetting LBQS 2212-1759 covering ∼0.2 deg 2 . The cluster identification is based on deep imaging with the ESO VLT and from the CFHT legacy survey, among others. The confirmation of cluster candidates is done by VLT/FORS2 multi-object spectroscopy. Photometric redshifts from the CFHTLS D4 were utilised to confirm the effectiveness of the X-ray cluster selection method. The survey sensitivity was computed with extensive Monte-Carlo simulations. Results. At a flux limit of S 0.5−2.0 keV ∼ 2.5 × 10 −15 erg s −1 we achieve a completeness level higher than 50% in an area of ∼0.13 deg 2 . We detect six galaxy clusters above this limit with optical counterparts, of which 5 are new spectroscopic discoveries. Two newly discovered X-ray luminous galaxy clusters are at z ≥ 1.0, another two at z = 0.41, and one at z = 0.34. For the most distant X-ray selected cluster in this field at z = 1.45, we find additional (active) member galaxies from both X-ray and spectroscopic data. Additionally, we find evidence of large-scale structures at moderate redshifts of z = 0.41 and z = 0.34. Conclusions. The quest for distant clusters in archival XMM-Newton data has led to detection of six clusters in a single field, making XMM-Newton an outstanding tool for cluster surveys. Three of these clusters are at z ≥ 1, which emphasises the valuable contribution of small, yet deep surveys to cosmology. Beta models are appropriate descriptions of the cluster surface brightness when performing cluster detection simulations to compute the X-ray selection function. The constructed log N − log S tends to favour a scenario where no evolution in the cluster X-ray luminosity function (XLF) takes place.