Claude Code stops billing server-side auto mode checks

In auto mode a classifier checks actions such as shell commands and network requests before they run. Anthropic's documentation says that from 2.1.278 Claude Code asks the server to perform those checks as part of the session's own model requests, and "doesn't charge for them when the server performs them." Server-side checks are on by default on the Anthropic API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry. Pro, Max and Team plans never see the dialog. When the server's checks stop reaching a session, Claude Code falls back to its own classifier requests, billed as before, and holds the next checked action behind a message reading "Auto mode is now free. This session isn't set up for that yet." A new Auto mode server row in /status reports which side of that line the current session is on.

The documented cause is an LLM gateway or proxy that strips or rewrites request headers, drops request fields it does not recognise, or edits responses by rewriting IDs or dropping keys from streaming events. Outside Amazon Bedrock sessions running a model older than Claude Opus 4.8, Anthropic says the dialog should appear only in that case. A team that already knows its gateway cannot carry the checks can set CLAUDE_CODE_AUTO_MODE_SERVER=0 to suppress the dialog and keep the billed behaviour. The documentation calls that variable "a temporary measure that exists to avoid disrupting existing workflows" and says Anthropic intends to deprecate and then remove it.

Why it matters: A gateway that rewrites requests or responses can now add classifier charges to every auto mode session. The gateway operator must preserve the server-side checks for users to receive the billing change. Teams can check the new /status row before assuming their fleet is using the free server-side classifier.