Claude Code v2.1.205: /doctor Becomes a Full Checkup, New Auto Mode Safeguards, and Agent View Improvements

Claude Code v2.1.205 combines a handful of feature changes with a broad set of fixes. The main themes are an expanded /doctor command, new auto mode safeguards around transcript files and destructive shell commands, and a more informative agent view in claude agents. The remaining changes fix issues with CLI flags, background agents, Windows, MCP, and plugins.

What’s New

`/doctor` Is Now a Full Setup Checkup

/doctor is now a full setup checkup that can diagnose and fix issues. /checkup works as an alias for it.

Auto Mode Safeguards

Auto mode has a new rule that blocks tampering with session transcript files, and it now asks before running rm -rf on a variable it can’t resolve from context. Separately, background task notifications now explicitly state that no human input has occurred, so fabricated in-transcript approvals are not acted on.

Agent View Improvements

Sessions that edit, merge, comment on, or push to an existing PR now link that PR in claude agents. Rows show a colored state word and a classifier-written headline instead of raw tool call text, and the peek opens with the full status, including the exact ask for blocked sessions.

Lower Memory Use During Auto-Updates

Auto-update binary downloads now stream to disk instead of buffering in memory, cutting the updater’s peak memory usage by roughly 400 MB.

Reserved MCP Server Names

The “Claude Browser” MCP server name is now reserved alongside “Claude Preview”, ahead of the Claude Desktop pane rename. User-configured MCP servers can no longer register under either name.

Fixes and Improvements

CLI flags: --json-schema no longer silently produces unstructured output when the schema is invalid, and schemas that use the format keyword are no longer rejected. A message sent while Claude was working is no longer silently lost when the turn ends at the --max-turns limit.
Background agents: Agents resumed with SendMessage no longer stay shown as “failed” or “completed” in the agent list, and jobs no longer flip from “needs input” back to “working” when the agent’s turn contains no readable text. claude attach now waits for a background agent that is in the middle of an upgrade restart instead of erroring.
Agent view: The view no longer renders one line too high and clips its header when the job list slightly overflows the screen. Session-to-PR linking no longer misses a PR created in a Bash call whose output exceeded the 30K inline limit.
Remote Control: Background tasks in the web and mobile Remote Control panels no longer show a stale “Running” status, because full task state is now forwarded on every membership change.
Windows: Worktree removal no longer deletes files outside the worktree when an NTFS junction or directory symlink exists inside it. Claude also no longer crashes when the directory it was launched from is deleted, locked, or unmounted while a command is running.
MCP: claude mcp add-from-claude-desktop no longer gets stuck when a server name contains unsupported characters; invalid names are now reported and the remaining servers still import.
Plugins: A plugin LSP server that fails to initialize no longer prevents a valid LSP server from another plugin from handling the same file extension.
Skills: Project verify skills are no longer rewritten on every session, only when a documented command changes.
Stability: Fixed a crash when a file watcher was closed while a directory scan was still in flight.
Cowork: VM-mode local-agent sessions no longer fail to start with “Not logged in · Please run /login” on CLI 2.1.203 and later.

How to Update

Check your installed version: claude --version
Update in place: claude update
If you configured your own MCP server named “Claude Browser” or “Claude Preview”, note that user-configured servers can no longer register under either name.