Claude Code v2.1.169 Major Updates

This is Ishikawa from the Cloud Business Division. Claude Code v2.1.169 (released 2026-06-08) has been released. This release focuses on new features such as a startup mode for troubleshooting and a working directory change command, along with fixes related to enterprise policy enforcement and security.

Update Summary

v2.1.169 includes 30 changes. The breakdown is 3 new features, 12 improvements, 12 fixes, 2 security updates, and 1 performance improvement. New features include --safe-mode, which launches with all customizations disabled, and the release is largely comprised of bug fixes related to Windows behavior, background sessions, and enterprise management settings. No breaking changes or deprecations are included.

Notable Updates

• New Feature: A --safe-mode flag (and CLAUDE_CODE_SAFE_MODE) has been added to launch with all customizations (CLAUDE.md, plugins, skills, hooks, MCP servers) disabled (v2.1.169). This can be used to isolate issues during troubleshooting. Additionally, a /cd command for changing the working directory and a disableBundledSkills setting for hiding bundled skills have also been added.
  
• Bug Fix: An issue where claude -p would slow down or appear to hang on Windows due to waiting for slash command/skill scans has been fixed (v2.1.169, a regression introduced in 2.1.161).
• Security: An issue where untrusted project settings could set OTEL client certificate paths without trust confirmation has been fixed (v2.1.169). Additionally, an issue where enterprise managed MCP policies (allowedMcpServers/deniedMcpServers) were not being applied in certain code paths has also been fixed.

Update Details

New Features

• A --safe-mode flag (and CLAUDE_CODE_SAFE_MODE environment variable) has been added to launch Claude Code with all customizations (CLAUDE.md, plugins, skills, hooks, MCP servers) disabled. This is useful for isolating whether a bug is caused by settings or extensions.
• A /cd command has been added to change the working directory mid-session without breaking the prompt cache.
• A disableBundledSkills setting and CLAUDE_CODE_DISABLE_BUNDLED_SKILLS environment variable have been added to hide bundled skills, workflows, and built-in slash commands from the model.

Security

• An issue where untrusted project settings could set OTEL client certificate paths without trust confirmation has been fixed.
• An issue where enterprise managed MCP policies (allowedMcpServers/deniedMcpServers) were not being applied during reconnection, IDE-entered settings, --mcp-config servers during the first session after installation, and before remote settings were loaded has been fixed. Additionally, cold start latency for organizations without remote settings has also been improved.

Improvements

• /workflows now opens immediately even while a turn is in progress.
• TaskCreate reliability has been improved. Invalid input is now automatically repaired, and validation errors for unloaded tools now include the schema.
• The error message displayed when an organization has disabled API key authentication has been improved, and guidance is now shown based on the source of the API key being used.
• The default 5-minute idle timeout for Vertex/Foundry has been restored, so stalled streams will be interrupted instead of hanging indefinitely. You can opt out with API_FORCE_IDLE_TIMEOUT=0.
• Even if remote managed settings contain invalid entries, the remaining valid policies are now applied and validation errors are displayed. The entire payload is no longer silently discarded as before.
• Background sessions now retain flags such as --ide, --chrome, --bare, and --remote-control across retire→wake cycles, and state validation on respawn has been strengthened.
• Background sessions are now informed that edits to shared checkouts are blocked until entering a worktree, avoiding unnecessary edit rejections before EnterWorktree.
• The threshold for the "CLAUDE.md is too long" warning now scales with the model's context window.
• The Windows auto-updater now stops retrying within a session if claude.exe is being held by another process.
• Color contrast for skill tags in the slash command menu has been improved.
• Apple/Google billing subscribers without a registered payment method are now guided on where to add one when claiming promotional credits.
• A hint suggesting claude agents when multiple sessions are running simultaneously has been added.

Performance

• CPU usage during response streaming and spinner animation has been reduced.

Fixes

• Fixed up/down arrow key history navigation behavior: Fixed an issue where the up/down arrow keys would skip over wrapped lines and move to command history on long input lines. Navigation now moves by visual line, and history recall enters from the nearest end.
• Fixed UI stall on macOS: Fixed an approximately 30–50ms UI stall that occurred at the start of each turn for macOS users logged in with claude.ai credentials.
• Fixed claude -p hang on Windows: Fixed an issue where claude -p would slow down or appear to hang on Windows due to waiting for slash command/skill scans (a regression introduced in 2.1.161).
• Fixed Remote Control stuck in reconnecting: Fixed an issue where Remote Control would get stuck in a "reconnecting" state when an OAuth token refresh occurred simultaneously during session resumption.
• Fixed re-display of stale permission prompts: Fixed an issue where stale prompts would be re-displayed every time a remote session was reconnected to where a worker had stopped while waiting for a permission/dialog prompt.
• Fixed missing sessions in claude agents --json: Fixed an issue where background sessions dispatched while blocked and immediately after were missing. An --all option to include completed sessions and new id and state fields have also been added.
• Fixed background agent ignoring environment variables: Fixed an issue where background agents dispatched to pre-warmed workers would ignore env values from project-level settings (e.g., ANTHROPIC_MODEL).
• In addition, numerous minor bugs have been fixed, including Git Credential Manager popups on Windows, footer hints when using custom status lines, agents view display corruption in WSL (Windows Terminal), MCPB plugin cache on Windows, and accumulation of .in_use lock files for plugins.

Closing

v2.1.169 is a release centered on the addition of new features including --safe-mode for troubleshooting, along with fixes related to enterprise management settings, security, and Windows/background session issues. Since no breaking changes or deprecations are included, this is an update that is easy to assess for impact. Those using claude -p in Windows environments or operating enterprise managed policies will particularly benefit from this update. Why not update and give it a try?

Recent Updates

https://dev.classmethod.jp/articles/20260524-claude-code-updates-v2-1-150/

https://dev.classmethod.jp/articles/20260524-claude-code-updates-v2-1-152/

https://dev.classmethod.jp/articles/20260528-claude-code-updates-v2-1-153/

https://dev.classmethod.jp/articles/20260529-cc-updates-v2-1-158/

https://dev.classmethod.jp/articles/20260602-cc-updates-v2-1-160/

https://dev.classmethod.jp/articles/20260603-cc-updates-v2-1-161/

https://dev.classmethod.jp/articles/20260603-cc-updates-v2-1-165/

https://dev.classmethod.jp/articles/20260608-cc-updates-v2-1-168/

References

https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md

https://www.npmjs.com/package/@anthropic-ai/claude-code