An architect is designing an agentic loop using the Claude API. The loop needs to handle tool calls, detect…

Question

An architect is designing an agentic loop using the Claude API. The loop needs to handle tool calls, detect completion, and manage errors. Which combination of stop_reason values correctly identifies when to continue the loop versus when to exit?

Accepted Answer

D. Continue on 'tool_use'; exit on 'end_turn', 'max_tokens', 'refusal', or 'stop_sequence'

Answer

A. Continue on 'tool_use' or 'end_turn'; exit on 'max_tokens' or 'refusal'

Answer

B. Continue on 'tool_use' or 'max_tokens'; exit on 'end_turn' or 'stop_sequence'

Answer

C. Continue on 'tool_use'; exit only on 'end_turn'

Question 196