Claude Code v2.1.207: Auto Mode Goes Default and Terminal Freezing Gets Fixed

Claude Code v2.1.207 drops the opt-in flag for auto mode and patches a security consent bypass that silently recorded remote managed settings as approved.

The release, tagged v2.1.207 on GitHub, bundles nine fixes and one feature change. The highlights:

• Auto mode on by default — available on Bedrock, Vertex AI, and Foundry without CLAUDE_CODE_ENABLE_AUTO_MODE. Disable via disableAutoMode in settings.
• Terminal freeze fix — streaming responses with long lists, tables, paragraphs, or code blocks no longer lock the terminal or lag keystrokes.
• Security consent bypass patched — remote managed settings from non-interactive runs (claude -p, the SDK) were being permanently recorded as consented without ever showing the security dialog.
• Launcher overwrite fixed — the auto-updater was clobbering custom launcher scripts or symlinks at ~/.local/bin/claude on every release. /doctor now reports an externally managed launcher.
• Spurious injection warnings silenced — benign system-generated conversation updates were triggering false prompt-injection warnings.
• Git config cleanup — extensions.worktreeConfig left behind in .git/config is now removed properly.

The security consent fix is the one to care about. Running claude -p or the SDK in a pipeline meant remote managed settings could be silently marked as approved — no dialog, no prompt, no audit trail. If you run Claude Code in CI or automation, this was a real exposure, not a cosmetic bug.

The terminal freeze is the quality-of-life fix anyone who has streamed a long code block has hit. Keystrokes lagging mid-response made interactive sessions feel broken.

Auto mode is live now across all three cloud providers. The terminal and security fixes ship with the update — run claude update or let the auto-updater handle it. If you had a custom launcher at ~/.local/bin/claude, check /doctor to confirm it’s intact.