Today, integrating AI-enabled algorithmic trading and market prediction services into different financial ecosystems is a high-priority requirement. The design of a reference architecture for these systems is important. Although research exists in this domain, a reliable and comprehensive reference architecture for financial market is needed. This paper proposes a new comprehensive reference architecture for trading systems (RATS) which considers all compulsive components in an auto trading system. The inclusion of AI has been considered as the main component of the RATS. The proposed architecture is described using a combination of a 4+1 View and UML diagrams. By evaluating the proposed reference architecture, the abstraction level has been kept as high as possible. The Architecture Tradeoff Analysis Method (ATAM) has been used to evaluate RATS’s quality attributes, including availability and operational consistency. The results confirm that the RATS effectively covered system attributes such as performance, interoperability, and modifiability for the auto trading systems in the financial market.