Plan where an agent should think hard and where it should not, turn by turn. Reasoning: medium. The schedule is a plan, not a computation. API only: configuration_update is an Astra feature of the Responses API in standard single-agent mode, not a ChatGPT control; Microsoft Learn lists it as unsupported on Azure OpenAI as of 6 September 2026. Set the first turn with reasoning.effort as usual. Free template — copy and paste into ChatGPT, Claude, or Copilot.