Anthropic: Claude Code v2.1.169 brings safe-mode flag, new /cd command, and key fixes
Anthropic has released Claude Code v2.1.169, an update to the CLI tool for agentic coding. The release introduces the --safe-mode flag for isolated troubleshooting, a new /cd command for changing the working directory without breaking the prompt cache, and fixes for enterprise MCP policies, UI freezes, and Windows issues.
This article was generated using artificial intelligence from primary sources.
Anthropic released Claude Code v2.1.169 on June 8, 2026, an update to the CLI tool for agentic coding. The version introduces new diagnostic controls, navigation enhancements, and a set of stability fixes for macOS and Windows platforms.
Safe-mode: environment isolation for diagnostics
The central new feature is the --safe-mode flag (or the equivalent env variable CLAUDE_CODE_SAFE_MODE). Safe-mode is a special startup profile that disables all CLAUDE.md configuration files, plugins, skills, hooks, and MCP servers at session start. Unlike a standard launch — where Claude Code automatically loads all configuration layers from the working directory and parent folders — a safe-mode session starts the CLI in a clean state, with no external extensions at all. Its purpose is purely troubleshooting: an administrator or developer can determine with precision whether a problem originates from custom configuration or from the tool’s core.
New /cd command and disableBundledSkills setting
Version 2.1.169 also introduces the /cd command, which changes the working directory mid-session without breaking the prompt cache. Previous versions did not support changing directories within a session — users had to shut down and restart the CLI, invalidating the accumulated conversation context and prompt cache. The new disableBundledSkills setting (and its corresponding CLAUDE_CODE_DISABLE_BUNDLED_SKILLS env variable) hides all built-in slash commands and skills from the model core, which is useful in enterprise environments with a narrow set of permitted operations.
What stability issues were resolved in v2.1.169?
This release fixes three separate bugs. The 30–50 ms UI freeze that appeared at the start of every turn for macOS OAuth authentication users has been eliminated. The hang of claude -p on Windows — a regression introduced in v2.1.161 — has been resolved. Additionally, enterprise MCP policies (allowedMcpServers / deniedMcpServers) were not being applied on reconnect or immediately after first installation; in v2.1.169 they are applied consistently. An --all flag was also added to claude agents --json, which now displays blocked and dispatched background sessions along with new id and state fields in the JSON output.
Claude Code v2.1.169 expands the tool’s diagnostic capabilities and closes stability gaps specific to macOS and Windows.
Frequently Asked Questions
- What does the --safe-mode flag do in Claude Code?
- The --safe-mode flag (or the CLAUDE_CODE_SAFE_MODE env variable) starts a session with all CLAUDE.md files, plugins, skills, hooks, and MCP servers fully disabled — intended for troubleshooting without external configuration influences.
- What fixes are included in v2.1.169?
- Three issues were fixed: a 30–50 ms UI freeze for macOS OAuth users, a hang in 'claude -p' on Windows (a regression from v2.1.161), and enterprise MCP policies not being applied on reconnect or immediately after first installation.
📬 AI news in your inbox
A daily digest built your way — pick topics, sources and cadence. One-click unsubscribe.