Source: github.com/anthropics/claude-code/releases/tag/v2.1.328
Interactive sessions that have no permission mode set now start in auto mode for Claude Opus 4.6 and Sonnet 4.6 on Amazon Bedrock, Google Cloud's Agent Platform, Microsoft Foundry and the Claude apps gateway. Plan mode and Bash prompts on third-party providers now offer auto mode as well, and disableAutoMode turns both off.
The mcp_tool hooks now accept "onFailure": "block", so a hook whose call fails blocks the action, matching command and HTTP hooks. Code Mode no longer runs UserPromptSubmit hooks on script output, and redaction of that output should use a PostToolUse hook with updatedToolOutput.
The /code-review command now reports only bugs and CLAUDE.md violations on most models and marks each finding Important, Pre-existing or Nit. Reuse and cleanup checks moved to /simplify. The release also adds self-hosted runner flags such as --repository-folder and --hook-label.
Why it matters if you build agents
Agents that run on these cloud platforms will now default to auto mode for the listed models, so permission handling in automated workflows may change.
Code that depends on UserPromptSubmit hooks seeing Code Mode output, or on managed PostToolUse hooks firing for refused calls, needs to be checked against the new behavior.
Get stories like this every weekday
One short email with the day's agent tooling news.