Recipe Overview

Some problems can be handled by one versatile agent. In a single-agent system, one LLM is given all tools and instructions to complete the task end-to-end. The problem it solves is simplicity: small tasks may not need coordination. For example, a single chatbot agent could both reason about a query and call a search API if needed. This works well for well-defined problems where the scope is manageable. The advantage is reduced complexity and easier debugging, while the limitation is scalability to very large or specialized tasks that might benefit from division of labor.