Single-agent agentic RAG: the router

The simplest architecture. One agent handles everything: it receives the query, decides where to look, retrieves, and integrates the answer. Its defining behavior is routing (the workflow pattern from Chapter 2).

The workflow: a coordinating agent analyzes the query, then selects the most suitable knowledge source from its options, which might include a structured database (via a Text-to-SQL engine), semantic search over unstructured documents, web search for real-time information, or a recommendation system. It retrieves from the chosen source, the LLM synthesizes the result, and the answer is returned.

Strengths: centralized simplicity (easy to design, implement, and maintain), efficiency (fewer agents, less coordination, lower resource use), dynamic routing in real time, and versatility across tools. It is ideal for simpler systems with a limited number of tools or data sources.

The survey's example is a customer support query about delivery status: the single agent pulls tracking details from the order database, fetches updates from the shipping API, and optionally checks the web for local conditions, then synthesizes one answer. This is the right starting architecture when one well-chosen source per query is enough.