Single Agent Architecture (SAA) is an artificial intelligence system architecture that refers to an execution mode in which a single intelligent agent independently completes the entire process of task understanding, reasoning and decision-making, step planning, and tool invocation. With the development of generative artificial intelligence and intelligent agent technology, this concept gradually took shape and became widely used in academia and industry between 2022 and 2023.

In a single-agent architecture, the system is typically driven by a core agent based on a large language model. Given a task and contextual information, this agent can autonomously analyze the problem and decompose the task, and invoke external tools or function interfaces to complete specific operations when necessary. The entire process is uniformly controlled by a single model, without relying on the cooperation of other agents, but it can perform a certain degree of self-correction through human feedback or internal mechanisms.