The Reasoning tool is the first—and most frequent—step the assistant takes after you send a message.
Think of it as the AI’s “inner monologue,” surfaced in plain language so you always know why it is about to do something.

​
🚀 Why it matters

  • Clarity & Trust – You see the numbered thought process before any agent or app acts on your data.
  • Zero-guess Workflow – The assistant pauses to ask clarifying questions before it locks in a plan.
  • Seamless Orchestration – Between every agent hand-off, Reasoning pops up again with a quick 1-2-sentence checkpoint, so the whole swarm stays in sync.

​
🔄 When it runs

  1. Immediately after every user message (mandatory first tool call).
  2. At least twice before the assistant generates an Agent Plan.
  3. Between every agent hand-off during multi-agent workflows.
  4. Even on tiny replies—no exceptions.