Introduction: In recent years, enormous advances have been made in genetic engineering of effector immune cells for cancer therapy. Although chimeric antigen receptors (CARs) have been widely used to redirect autologous T-cell specificity against hematological malignancies, yielding impressive clinical results, studies with CAR-modified natural killer (NK) cells are still restricted to pre-clinical and some clinical studies. Genetic engineering is an important tool for redirecting the function of various types of immune cells and their use for therapeutic purposes. One of the major obstacles for NK cell immunotherapy is the lack of an efficient method for gene transfer. Lentiviral vectors have been proven to be a safe tool for genetic engineering, however, lentiviral transduction is inefficient for NK cells. This study aims to develop a non-viral episomal vector for genetic modification of NK cells. Methods: The S/MAR-based episomal vector pEPI-eGFP supplied the basic functional architecture for the reprogramming of our target cells. The system contains transcriptional units optimized for mammalian expression such as the replication-Initiation Region (IR) from the β-globin locus (for plasmid replication), a scaffold/matrix attachment region sequence (S/MARS) for persistent transcription, and an SFFV promoter to drive the expression of our CAR component. Based on this vector, we have cloned a second-generation anti-CD19 CAR (41BB-CD3z) and two fourth-generation anti-CD19 CARs (41BB-CD3z- IL15 and 2B4-CD3z-IL15). Results and discussion: Firstly, we evaluated the functionality of our episomal pEPIR vector. We observed a transfection efficiency of about 47% with our vector when transfected in HEK cells. Further, we assessed the stability of our construct using the same cell lineage, as the HEK cell line has a high replication rate. Our results show a stable expression (45.2%) for 90 days of culture and no cell growth disturbance due to the presence of the plasmid. These results prove that the sequences included in this vector, such as S/MAR and IR, allow it to replicate autonomously within the host cells, without loss of expression during cell passages. These results are very promising, as the expression of a common episomal vector lasts only about 1–2 weeks. Also, we have cloned the CAR molecules into this vector, resulting in the following constructs: pepIR-CD19.CAR-41BB-CD3z-GFP, pepIR-CD19.CAR-41BB-CD3z-IL-15 and pepIR-CD19.CAR-2B4-CD3z-IL-15. Conclusions: Our episomal vector supported stable transfections and the IR element together with S/MAR contribute to the formation of autonomously replicating genetic elements. The next steps include the evaluation of the transfection efficiency of NK cells and the expression stability within them, as well as to test the in vitro and in vivo function of non-viral CAR-NK cells. Financially supported by CAPES, FAPESP 2019/25309-0, CTC-Center for Cell-based Therapy (FAPESP 2013/08135-2) and National Institute of Science and Technology in Stem Cell and Cell Therapy (CNPq 573754-2008-0 and FAPESP 2008/578773).