In a multi-tool Claude agent, what should the application do after…

Question

In a multi-tool Claude agent, what should the application do after receiving a tooluse block?

Accepted Answer

Execute authorized tools, then return toolresult content so Claude can continue reasoning — tooluse means 'please call this.' You run it (if allowed), send toolresult back, and Claude continues.

Answer

Ignore it and always reply with empty content

Answer

Execute the tool client-side without authorization checks

Answer

Convert tooluse into a system prompt rewrite automatically

In a multi-tool Claude agent, what should the application do after receiving a tool_use block?

Short Explanation and Infographic

Full Explanation