Claude Code 2.1.304 adds auto mode setup and a Managed Agents security check

what changed

Claude Code 2.1.304 was published on 19 October 2026. These changes stand out:

• /auto-mode-setup is a guided setup that teaches auto mode about your environment.
• /claude-api managed-agents-secure compares a Managed Agents agent’s settings against Anthropic’s security guide. It does not change anything.
• Tool search is now on by default behind a custom ANTHROPIC_BASE_URL. If the gateway answers with a 400 or 422, Claude Code retries without it.
• Code Review gets review levels, Quick, Standard and Deep. They can be set per organization and per repository.

The release also lists 43 bug fixes.

why it matters

Teams that run Claude Code through an internal gateway get tool search without extra config, with a safe fallback if the gateway rejects it. That helps sessions with many MCP tools.

Teams that roll out auto mode get a guided way to describe their environment. The Managed Agents check gives a read-only review of an agent against Anthropic’s own guide.

what we would do

• Run /auto-mode-setup in one pilot repository and read what it records before you share the result with the team.
• Run the Managed Agents check on each agent you operate, and file the gaps it shows as normal tickets.
• If you use a gateway, test a session with many MCP tools and confirm the retry works with your gateway.
• Decide a default Code Review level per repository, and write down why.

Our enable service helps teams set these defaults.