Claude Code 2.1.326 adds Code Mode for agent teams and changes HIPAA auto-mode defaults
Claude Code 2.1.326 lets sessions with agent teams start in Code Mode and makes terminal and VS Code sessions in HIPAA-configured organizations start in auto mode.
Anthropic's Claude Code 2.1.326 changelog, dated November 16, 2026, makes two changes. Terminal and VS Code sessions in organizations with the Claude Code + Cowork HIPAA configuration now start in auto mode by default. Sessions with agent teams enabled start in Code Mode for users who already turned Code Mode on. The release notes describe both as default changes, so the details of those defaults matter.
The HIPAA change comes first. The changelog says these sessions start in auto mode by default, and the disableAutoMode setting turns this off. A related entry says terminal and VS Code sessions with no mode set now start in auto mode rather than Manual mode when the organization's policy has not loaded. The same entry says sessions that start in auto by default take a server-managed defaultMode once settings load. They fall back to Manual if settings fail with no stored copy and policy fails or requires it. So the effective starting mode depends on several conditions, not one switch. The changelog also says that, for organizations with the HIPAA configuration on a direct Anthropic API connection, the server-side classifier now checks actions, as it does for other organizations. The notes do not explain how that classifier works or what it blocks.
The Code Mode change is a separate item. Users who had already enabled Code Mode now find that sessions with agent teams on begin in Code Mode. The release also adds the ability to turn Code Mode on while agent teams are enabled. The changelog does not describe what Code Mode does beyond this, but its other entries show it has its own permission checks. Several items say permission checks in Code Mode now ask or refuse more sandboxed commands. One says a `$` shell helper in Code Mode refuses some values with an error, so nothing runs.
The release also includes other changes to permission and sandbox handling. Auto mode now stops some shell commands it used to run. It ignores some wildcard shell allow rules, so the safety check decides those commands, and it judges wildcard allow rules whose program is a Windows backslash path. The changelog also says a managed autoMode.environment list without "$defaults" replaces every built-in entry, including in Claude Desktop and Agent SDK apps. It further says a hook's "decision": "block" now holds in more cases, so a hook that always prints one may start blocking calls. Administrators who deploy managed settings should read those entries closely.
Two items concern networking and skills. Remote Control now sends its requests through a CONNECT tunnel, so a proxy that refuses CONNECT stops Remote Control with an error. The /claude-api skill now builds agents on Managed Agents by default unless the user wants to host the agent loop themselves.
The VS Code setting claudeCode.promptSuggestions turns prompt suggestions off. The notes list it under Added.
What changed
Claude Code 2.1.326 sessions in organizations with the Claude Code + Cowork HIPAA configuration now start in auto mode by default, and sessions with agent teams enabled start in Code Mode for users who had already turned Code Mode on.
Type: Breaking change · Takes effect: Monday, November 16, 2026 · Applies to: Claude Code 2.1.326, Claude Code + Cowork HIPAA configuration, Code Mode, agent teams, Remote Control, /claude-api, claudeCode.promptSuggestions
Why it matters
Organizations using the HIPAA configuration get a different default permission mode, and proxy setups that refuse CONNECT requests will stop Remote Control.
What you can do
Organizations under the Claude Code + Cowork HIPAA configuration should decide whether to keep auto mode by setting disableAutoMode, and check proxies before relying on Remote Control.
Original source
Where we saw it
Claude Code changelog
AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works