wab
wab — yIlaD.
{study notes}
- {Agent} - Claude'e autonomous version, messaging loop'e both sides running, human middle'e without. Task receives, tool picks, code executes loop'e until task done.
- {Agentic workflow} - Act, result look, next decide, continue pattern. Single API call only one response returns; automation needs observe-decide-act repeat.
- {Agent loop} - Messages Claude send tools available. Claude final answer or tool request responds. Code tool executes. Result Claude send back. {Stop reason} "end_turn" until repeat.
- {Tools array} - Claude available what tells. Name, description, JSON schema inputs'e. Claude knows what use can.
- {Tool use} - Claude tool requests when, code executes. Result back messages push, loop again Claude answer can.
- {Stop reason} - "end_turn" - Claude done, answer final give. "tooluse" - tool execute, continue loop.
- Production scaling - Same loop shape mock weather demo from compliance agent production to. Tools real, plumbing different only. Observe-decide-act-repeat identical stays.
{flashcards} 8 {cards}
{question}
yI'ang · ←/→
{answer}
{flip}
{knowledge check} 6 {questions}