Anthropic released Claude Code v2.1.234, introducing several new features alongside significant bug fixes and security improvements. Key additions include a GitLab merge request badge, automatic session continuation after usage limits reset, and an optional environment variable for customizing transcript directory names.

  • Added the `CLAUDE_CODE_PROJECT_DIR_NAME` environment variable to customize per-project transcript directories.
  • Introduced a `selection:clear` keybinding action for clearing in-app text selections.
  • Displayed GitLab merge request badges in the footer and statusline for authenticated repositories.
  • Enabled automatic session continuation when claude.ai usage limits reset, configurable via `/config`.
  • Hardened security by rejecting Windows NT-namespace (`\??\`) paths to prevent credential leaks.
  • Fixed issues with auto mode network access checks, session-scoped permission drops, and markdown rendering performance.
  • Improved transcript markdown rendering, error messages for malformed API responses, and auto-generated session titles.
  • Reduced context cost of the `claude-api` skill from ~200k+ tokens to ~25k by loading reference docs on demand.

These changes enhance usability through better GitLab integration and session management while improving security posture against specific credential-leak vectors.