Claude Certified Developer (CCD)

Context: An architect is building a multi-step agent that uses tools. The agentic loop must continue executing as long as Claude requests tool calls but should terminate immediately once Claude provides a final, conclusive answer to the user. Problem: The loop occasionally terminates prematurely before all sub-tool calls are processed. Question: Which `stop_reason` value in the Messages API response definitively indicates that Claude has finished its turn and is NOT waiting for a tool result, signaling that the loop should end?