The BRAC MANOSHI programme established BRAC delivery centres (BDCs) across slums in urban areas of Bangladesh with the intention to reduce unsafe delivery at home with an affordable cost. Subsequently, 27 delivery centres were upgraded to BRAC Maternity Centres (BMCs) with added capacity of midwifery services under supervision of MBBS doctors. These structural changes led to a reduction in referral of various delivery cases. Thus, it is important to investigate the contribution of BMC and BDC in the poor urban setup of Bangladesh in improving maternal and neonatal healthcare services, their referral compliance and cost of delivery. A mixed method exploratory study was conducted in 10 branch offices of MANOSHI in urban areas of Dhaka, Sylhet and Chittagong during March to June. Out of 45 branch offices in three city corporations 10 were selected randomly. Approximately 2500 households (HHs) surrounding each of the selected BMC or BDC were selected in the sampling frame for enumeration. A total of 25,700 HHs were enumerated, out of which 6878 eligible mothers were identified. Finally, 1100 women were interviewed randomly for the main survey. Findings revealed that ANC checkups by BRAC SK improved institutional delivery significantly (P=0.000). However, ANC checkups by any provider did not confirm the provision of quality of care, as not all necessary examinations for identification of risk pregnancy were performed. About 39% of the deliveries were conducted at home with the assistance of TTBA or TBA and 16% of them were conducted either at BMC or at BDC. The median delivery cost at BMC was BDT 1330/- (IQR: 1200-1790) and at BDC BDT 480/- (IQR: 200-822) which was significantly lower than that of home (BDT 950/-), public (BDT 5230/-), private (BDT 19700/-) and other NGO facilities (BDT 5100/-). We observed that poverty and illiteracy were major barriers to skilled assistance at delivery, while poverty was reported as a barrier to referral compliance. Unless, they suffered from any delivery complication these women hardly sought treatment from professional care providers. The referral system of BMC and BDC was functioning specially focusing on cases related to prolonged labour, obstructed labor and malposition of the fetus. Differential coverage of different components of essential newborn care was observed in different birthplaces. This study suggests that, the MANOSHI programme has room for improving neonatal and maternal health service delivery in poor urban communities through BMC and BDC. Proper training for staff of BMC and BDC and up-gradation of both the facilities to basic EmOC would enhance their capacity to handle the complicated delivery cases.