Claude Code 2.1.161: Parallel Tool Calls, Metrics Labels, Agent Fixes

2.1.161 brings parallel tool resilience, OTEL metric labels for custom dimensions, and fixes for agents, sessions, and accessibility. Parallel Bash calls now fail independently.

Claude Code 2.1.161: Parallel Tool Calls, Metrics Labels, Agent Fixes

TL;DR

  • Parallel tool calls now fail independently — a broken Bash command won't cancel other calls in the batch
  • OTEL metrics now include custom dimension labels (team, repo, etc.) for better usage slicing
  • Multiple agent and session fixes, including worktree support and background subagent output corruption

New

  • OTEL metric labelsOTEL_RESOURCE_ATTRIBUTES values now appear as labels on datapoints, letting you slice usage by custom dimensions
  • Agent progress displayclaude agents rows show done/total before details when work fans out; peek shows the longest-running item
  • MCP connector filtering/mcp now collapses unused claude.ai connectors behind a "Show unused connectors" row
  • Parallel tool resilience — failed Bash commands no longer cancel other tool calls in the same batch; each returns its own result independently
  • Fullscreen clipboard improvements — Linux now uses wl-copy/xclip/xsel when available, copies to both clipboard and PRIMARY selection for middle-click paste, and shows the correct terminal-specific key hint

Fixed

  • Effort dialog, workflow animations, and prompt keyword shimmer now respect "Reduce motion" accessibility setting
  • Third-party provider sessions (Bedrock, Vertex, Foundry, Mantle) no longer blocked by forceLoginOrgUUID/forceLoginMethod managed policies (regression from 2.1.146)
  • Background subagent output corrupting claude -p stdout with --output-format text or json
  • /usage-credits no longer triggers re-login for Team/Enterprise admins; now points to org usage settings
  • /autofix-pr false "cannot run on default branch" error when session is in a git worktree or other repository
  • --resume picker now shows sessions from current directory in non-worktree contexts (e.g., jj workspaces)
  • Windows hooks invoking bash explicitly (e.g., /usr/bin/bash script.sh) failing with "command not found"
  • OpenTelemetry log events (user_prompt, api_request, tool_result, tool_decision) silently dropped before telemetry initialization
  • claude mcp list/get/add no longer prints secrets; ${VAR} references unexpanded, credential headers and URL secrets redacted
  • Workflow agents with isolation: "worktree" in background sessions blocked from editing files in their own worktree
  • Background sessions from claude agents booting on stale model from daemon environment instead of settings.json
  • Potential crash when rendering Write tool results after session resume
  • Completed subagents stuck showing as running when error occurs during result finalization
  • EADDRINUSE errors from tools binding Unix sockets under $TMPDIR with deep CLAUDE_CODE_TMPDIR paths
  • Terminal rendering performance improved by stabilizing layout engine JIT compilation
  • Large file write rendering performance improved
  • [VSCode] Added tip for disabling terminal GPU acceleration to fix garbled glyphs

Update Claude Code: brew upgrade claude-code or download from code.claude.com

Source: Claude Code