Why does Claude Code say my session isn't eligible for no-charge classifier requests?

The notice means the server's checks stopped reaching that session, so Claude Code 2.1.278 fell back to its own billed classifier requests. The usual cause is an LLM gateway or proxy that strips or rewrites request headers and fields; the notice names the gateway when Claude Code can identify one. Nothing breaks — auto mode keeps working and is billed the way it was before.

How do I turn the server-side classifier off?

Set CLAUDE_CODE_AUTO_MODE_SERVER=0 before starting the session, in your shell or the env settings key, and Claude Code always uses its own classifier requests with no notice. Anthropic describes the variable as temporary and says it may be removed in a later release. It is not read on a direct connection to the Anthropic API.

Does this change affect Claude Pro, Max or Team plans?

No. Pro, Max and Team plans never see the eligibility notice, and Claude Code 2.1.278 asks for server-side checks by default only on Enterprise plans, accounts using the Claude API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry. Whether a given platform or region actually performs the checks depends on that platform's own rollout.

What does my gateway need to change to qualify?

A gateway has to pass requests and replies through unchanged: forward request headers and body fields it does not recognise, including the safeguards field, and return responses and streaming events without dropping keys such as safeguard_results or rewriting tool-use IDs. Anthropic's gateway compatibility guide covers the rules, and a gateway that follows them keeps working with later features too.

Anthropic · 2026-09-19 · major

Claude Code 2.1.278 — auto mode's safety checks stop costing you tokens

Claude Code 2.1.278 moves auto mode's safety classifier to the server and stops charging for it. Enterprise, Claude API, Bedrock, Google Cloud Agent Platform and Microsoft Foundry sessions ask for server-side checks by default.

Auto mode's pre-flight safety checks move to Anthropic's servers, and you are no longer billed for them.

Quick facts

Version	v2.1.278
Maker	Anthropic
Released	2026-09-19
What changed	Auto mode classifier checks run server-side, unbilled
On by default	Enterprise, Claude API, AWS, Bedrock, Agent Platform, Foundry
Never affected	Pro, Max and Team plans
Opt-out	CLAUDE_CODE_AUTO_MODE_SERVER=0

What is it?

Auto mode's classifier stops appearing on the bill in Claude Code 2.1.278. The classifier is the safety check that inspects an action — a shell command, a network request — before it runs, which is what lets auto mode skip the permission prompt. Until now each check was a separate billed model request; from this version Claude Code asks the server to fold those checks into the session's own requests and does not charge for them.

How does it work?

The request carries a safeguards field and the reply carries safeguard_results, so the check rides along with work the session was already doing. That only functions end to end when nothing in the path rewrites the traffic. An LLM gateway or proxy that strips unknown headers, drops unrecognised body fields or rewrites tool-use IDs breaks the handshake, and Claude Code falls back to its own billed classifier requests. A new Auto mode server row in /status reads Enabled or Disabled so you can see which side a session is on.

Why does it matter?

Teams that left auto mode switched off to avoid classifier overhead can now turn it on without a second line on the invoice — the setting is on by default for Enterprise plans and Claude API accounts, and on Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry, subject to each platform's rollout. Pro, Max and Team plans are unaffected. When the server cannot be reached the session pauses once with a notice explaining it is not eligible, and billing continues exactly as before.

Who is it for?

teams running Claude Code in auto mode on enterprise or cloud platforms

Frequently asked questions

Why does Claude Code say my session isn't eligible for no-charge classifier requests?: The notice means the server's checks stopped reaching that session, so Claude Code 2.1.278 fell back to its own billed classifier requests. The usual cause is an LLM gateway or proxy that strips or rewrites request headers and fields; the notice names the gateway when Claude Code can identify one. Nothing breaks — auto mode keeps working and is billed the way it was before.
How do I turn the server-side classifier off?: Set CLAUDE_CODE_AUTO_MODE_SERVER=0 before starting the session, in your shell or the env settings key, and Claude Code always uses its own classifier requests with no notice. Anthropic describes the variable as temporary and says it may be removed in a later release. It is not read on a direct connection to the Anthropic API.
Does this change affect Claude Pro, Max or Team plans?: No. Pro, Max and Team plans never see the eligibility notice, and Claude Code 2.1.278 asks for server-side checks by default only on Enterprise plans, accounts using the Claude API, Claude Platform on AWS, Amazon Bedrock, Google Cloud's Agent Platform and Microsoft Foundry. Whether a given platform or region actually performs the checks depends on that platform's own rollout.
What does my gateway need to change to qualify?: A gateway has to pass requests and replies through unchanged: forward request headers and body fields it does not recognise, including the safeguards field, and return responses and streaming events without dropping keys such as safeguard_results or rewriting tool-use IDs. Anthropic's gateway compatibility guide covers the rules, and a gateway that follows them keeps working with later features too.

Try it

npm i -g @anthropic-ai/claude-code@2.1.278