Claude Code v2.1.234: GitLab MR Badge, Auto-Continue, Security Hardening

Claude Code v2.1.234 is out, and it’s a solid patch release with a few notable additions and a long list of fixes. The headline features are a GitLab merge request status badge, automatic session continuation after usage-limit resets, and a security hardening against NTLM credential-leak vectors.

Key additions

• GitLab MR badge: If your repo has a GitLab remote and you have the glab CLI authenticated, the footer and statusline now show a badge with the MR number and its state (draft, pending, or green).
• Auto-continue at usage limit: When your claude.ai usage limit resets, Claude Code automatically resumes your session. You can disable this in /config under “Continue automatically at usage limit”.
• CLAUDE_CODE_PROJECT_DIR_NAME env var: Lets hosts that give each session its own config directory choose a short name for the per-project transcript directory.
• selection:clear keybinding action: Bind a key to clear an in-app text selection, including in the agents view.

Security fix

The release hardens remote file reads, session restore, CLAUDE.md includes, workflow scripts, and file uploads against Windows NT-namespace paths (\??\). This closes the remaining pre-approval file-access vector that could be exploited for NTLM credential leaks.

Fixes worth noting

• Auto mode in very long sessions repeatedly re-checking and denying network access for sandboxed commands after compaction has been fixed.
• Session-scoped permission answers (including denies) are no longer dropped when answering background subagent tool permission prompts.
• Fixed a crash on the non-streaming fallback path when a thinking block or text block was missing its required field.
• Markdown rendering performance for unusual Unicode sequences is improved.
• SendMessage no longer rejects recipients copied from ListAgents when the session name is near the 200-character cap or emoji-heavy.
• Git remote parsing with unusual userinfo now correctly identifies the host, avoiding wrong-host links and behavior.
• MCP diagnostics no longer print resolved secrets: scope-conflict warnings show the configured ${VAR} form, and connection-failure details show only the server origin.
• strictKnownMarketplaces now rejects SCP-style git marketplace sources whose host differs from what git would actually connect to.
• Fullscreen copy of modal text (like the /login OAuth URL) no longer loses characters.
• Various UI/UX fixes: horizontal rules in markdown, consecutive shell commands splitting into separate rows, dialogs dismissed by shell command completion, queued ! commands being sent as plain text after editing, and permission-mode loss when accepting the fullscreen renderer prompt.
• /tui now declines to switch if it would drop launch --allowed-tools / --disallowed-tools rules.
• Credential masking on relayed permission previews can no longer hide commands, paths, or destinations from the approver.

Full changelog is on GitHub. If you’re on a recent version, upgrading is straightforward.