What probably matters to youSection of the release
What
An LLM gateway is a server you put between Claude Code and the model provider. Claude Code's bundled guidance for gateways now says that when a gateway replies with a 501 not_supported error, it should also send the header x-should-retry: false. Without that header, Claude Code tries the request once more, about half a second later, before it falls back.
Why
If you run a gateway, sending this header saves a pointless retry and a short delay whenever an endpoint is not available.
