The Pulse
Claude Code 2.1.278 Makes Server-Side Auto-Mode Classification the Default
Claude Code v2.1.278 changes auto mode for Claude API, Enterprise, Bedrock, Vertex, Foundry and gateway deployments, adds a status indicator and warns on billed fallback.
AI.info Team ·
Claude Code v2.1.278 changes auto mode to use the server-side classifier by default for Claude API and Enterprise users, as well as deployments on Amazon Bedrock, Google Vertex, Microsoft Foundry and gateways. GitHub lists the release as “19 Sep 03:10”; the release page does not state a year. The release notes say the server-side classifier does not charge for classifier overhead.
“Changed auto mode for Claude API and Enterprise users, and on Bedrock, Vertex, Foundry and gateways, to default to the server-side classifier, which does not charge for classifier overhead (
— ashwin-ant, GitHub release publisherCLAUDE_CODE_AUTO_MODE_SERVER=0opts out on Bedrock, Vertex, Foundry and gateways); warns on billed fallback.”
Server-Side Checks Become the Default
The change covers Claude API and Enterprise users, along with users running Claude Code through Bedrock, Vertex, Foundry and gateways. In those environments, auto mode now uses the server-side classifier by default.
The release notes do not describe a new permission model or provide details about how the classifier evaluates individual actions. They identify the change as a shift in where auto-mode classification runs and specify that classifier overhead is not charged.
For Bedrock, Vertex, Foundry and gateway deployments, the release note says that CLAUDE_CODE_AUTO_MODE_SERVER=0 opts out of the server-side classifier. It does not describe the variable as a permanent configuration or state when it might be removed. The note also does not say that version 2.1.278 newly introduced the variable.
Billed Fallback Warning
Claude Code v2.1.278 also warns on billed fallback. The release notes do not give further details about the conditions that trigger fallback, the wording of the warning or the action a user should take after seeing it. They establish only that the release warns users when auto mode falls back to a billed classifier path.
The release links to documentation about auto-mode classifier billing, but the release entry itself does not explain the fallback process or identify the specific classifier used after a fallback.
/status Shows the Classifier Location
The release adds an Auto mode server row to the /status command. According to Anthropic’s note, the row shows whether this session’s auto-mode classifier runs on the server.
This gives developers a way to check the classifier location from within the session instead of relying only on deployment configuration. The release does not specify the exact status values displayed in the row.
Supported Deployment Names
Anthropic’s release names Google Vertex, not Google Cloud’s Agent Platform. The complete list in the release note is Claude API, Enterprise, Bedrock, Vertex, Foundry and gateways. Version 2.1.278 makes server-side auto-mode classification the default across those listed environments, states that CLAUDE_CODE_AUTO_MODE_SERVER=0 opts out on Bedrock, Vertex, Foundry and gateways, warns about billed fallback and exposes the server status through /status.