A fintech startup called PayFlow is building an AI customer support agent that handles account inquiries, transaction disputes, and fraud reports. The engineering team is evaluating how to structure the agent using the Agents SDK. They need the agent to maintain conversation state across multiple turns, handle tool calls for checking account balances and transaction history, and survive brief network interruptions without losing context. The team creates a class that extends Agent from the SDK. What architectural benefit does extending Agent provide for PayFlow's use case?
Agent Class Lifecycle
Loading lesson content...
A developer creates a chatbot agent that extends Agent. They want the agent to maintain conversation history between turns. What mechanism should they use?
An agent needs to handle 10,000+ concurrent conversations. Each conversation uses Agent.state. What's the scaling consideration?