Introduction Conservative Dentistry has gradually replaced traditional invasive procedures with minimally invasive techniques that rely on adhesion to tooth substrates [1]. Due to these recent developments, indirect restorations are increasingly used in contemporary practice. In order to bond these restorations, a vast choice of materials exist, and these include flowable resin composites [2,3]. However, studies featuring these materials are seldom found [3]. The aim of this in vitro pilot study was to evaluate and compare the microleakage of indirect restorations luted with a flowable composite and different luting agents. Materials and methods This study was approved by the Ethics Committee of Egas Moniz, CRL. Thirty human molars were randomly divided between 3 groups (n = 10) according to the luting agent used: resin cement Bifix QM (VOCO GmbH, Cuxhaven, Germany) (G1), pre-heated resin composite Z100 MP Restorative (3 M ESPE) (G2) and flowable composite GrandioSO Flow (VOCO GmbH, Cuxhaven, Germany) (G3). Standardised class V cavities were prepared in the buccal surface and immediate dentine sealing was carried out with Optibond FL (Kerr). Composite restorations were made using a semi-direct technique with GrandioSO (VOCO GmbH, Cuxhaven, Germany). After 24 h cavities and restorations were pre-treated and adhesively luted according to the groups. After finishing and polishing, specimens were stored in distilled water at 37 °C for 24 h. After thermal aging (500 cycles at 5–55 °C) teeth were sealed and immersed in 0.5% basic fuchsine dye for 4 h. Each specimen was then sectioned vertically and microleakage was assessed and classified for both occlusal and cervical margins according to ISO/TS 11405:2015. The results were statistically analysed by Kruskal–Wallis (KW) test and binomial data analysis, at a significance level of 5% (SPSS 24.0). Results There were no significant differences in microleakage scores between the tested groups (p > .05, KW). All the groups scored more leakage in cervical margins. Binomial analysis confirmed that the success rate was material and margin dependent. The probability of failure of a cervical margin bonded to Z100 (G2) and to GrandioSO Flow (G3) using a semi-direct technique was significantly higher than 50% (p = .021) and (p = .002) respectively. Z100 had 100% failure rate in the cervical margin. Discussion and conclusions Longevity of a bonded interface is determined by sealing and microleakage. When this interface is compromised, failure of the restoration may happen [4]. Microleakage occurred in composite restorations made semi-directly the same way amongst the different materials used for bonding. However, cervical margins bonded with resin composites carry a greater chance of failure.