Claude Code 2.1.328 is the latest release in Anthropic’s public changelog. Two changes matter to anyone paying for AI-assisted development.
What changed
- Auto mode on cloud platforms. Auto mode is now available for Claude Opus 4.6 and Sonnet 4.6 on Amazon Bedrock, Google Cloud’s Agent Platform, Microsoft Foundry and the Claude apps gateway. Interactive sessions with no permission mode set start in it. In plain terms: the agent can proceed with fewer “are you sure?” pauses, on the cloud accounts large companies already use.
- A narrower code review.
/code-reviewon most models now reports only bugs and CLAUDE.md violations (the written rules a team gives the agent). Reuse and cleanup checks moved to a separate/simplifycommand. - Safer automation. Hooks, the automatic checks that run around the agent’s actions, can now block an action if the check itself fails to run.
- Housekeeping. The self-hosted runner gains new options, and there are fixes across Code Mode, terminal input, scheduled tasks and MCP connectors.
Why it matters
The pattern across recent releases is consistent: AI coding agents are being given more freedom, and more guardrails to match. Auto mode reduces interruptions. The hook change means a broken safety check stops work instead of being skipped. The review change reflects a lesson every buyer should absorb: a review that lists everything gets ignored, and one that lists real bugs gets read.
So what
If your supplier uses AI coding tools, ask which permission settings they run and what checks stop an agent acting on something it should not. The answer should be specific. Our AI-assisted development work pairs agent speed with human review and tests. If you are planning a build, start a project.