Claude Code

Claude Code 2.1.328: auto mode for Opus 4.6 and Sonnet 4.6 on Bedrock, Foundry and friends

2 min read AI-generated

Also new: /code-review now reports only bugs and CLAUDE.md violations. And Code Mode drops a 0.75-second pause after almost every script.

Featured image for "Claude Code 2.1.328: auto mode for Opus 4.6 and Sonnet 4.6 on Bedrock, Foundry and friends"

Claude Code 2.1.328 landed on npm on Wednesday evening at 23:09 UTC. The changelog is long, and most of it is bug fixes. Three changes will matter to a lot of teams, though.

Auto mode on the cloud platforms

If you run Claude Code through Amazon Bedrock, Google’s Agent Platform, Microsoft Foundry or the Claude apps gateway, you now get auto mode for Claude Opus 4.6 and Sonnet 4.6. Interactive sessions with no permission mode set start in it. Plan mode and the Bash prompts offer it too, the same way they do for newer models. If you’d rather not, disableAutoMode turns both off. Until now, auto mode on these platforms was limited to the newer models. I last covered how it behaves in Claude Desktop with desktop release 5.11492.

A stricter /code-review

On most models, /code-review now reports only bugs and CLAUDE.md violations. Each finding is tagged Important, Pre-existing or Nit, and every summary ends with a suggested fix. Cleanup work like reuse, simplification and efficiency moves to /simplify. A clean run now reads “Code review found no issues.”

Hooks that can block

mcp_tool hooks now accept "onFailure": "block", like command and HTTP hooks already did: if the call fails, the action is blocked. For admins, apps that embed Claude Code can now switch on the policy plugin’s allowManagedModsOnly option. An admin’s managed settings still win.

Code Mode gets faster

Code Mode, around since 2.1.323, gets a dozen or so fixes. In auto and bypass modes, a 0.75-second pause after most scripts is gone. You can now see which command a script is running and its output, and Claude starts dev servers there as background tasks that show up in /tasks. Self-hosted runners get new options such as --repository-folder and --hook-label, plus two new metrics.

Reviews get narrower

If your company buys Claude through its cloud provider and still runs Opus 4.6 or Sonnet 4.6 there, auto mode is the line to read. You get the feature without switching models. The direction for /code-review is just as clear: fewer comments, each one ranked. If you parse its output in a CI pipeline, you’ll need to update the format.

Sources

Claude CodeAuto ModeBedrockCode ModeRelease