ReleaseBytes
anthropic Claude Code Releases · · 2.1.210

Claude Code v2.1.210: Elapsed Time Counter, Permission Warnings, and Fixes

sdk
patch

Claude Code version 2.1.210 introduces a live elapsed-time counter for long-running tool calls, providing better visibility into progress. It also adds startup warnings for specific permission rules and resolves numerous bugs related to Git operations, plugin behavior, and UI rendering. This update aims to improve the user experience and stability for developers using Claude Code for code generation and editing tasks.

  • →Add live elapsed-time counter to collapsed tool summaries
  • →Add startup warnings for Write, NotebookEdit, and Glob permission rules
  • →Fix isolation: 'worktree' subagents running commands against main repo
  • →Fix ultracode keyword opt-in firing on non-human-originated input
  • →Fix rendered text fragment leaking into crash telemetry
Enhancements (2) ›
  • Add live elapsed-time counter to collapsed tool summaries

    A live elapsed-time counter has been added to the collapsed tool summary line. This allows users to see visible ticking for long-running tool calls, preventing them from appearing stuck.

  • Add startup warnings for Write, NotebookEdit, and Glob permission rules

    Startup warnings are now provided for permission rules using `Write(path)`, `NotebookEdit(path)`, and `Glob(path)`. Users are advised to use `Edit(path)` or `Read(path)` instead for better security and functionality.

Fixes (4) ›
  • Fix isolation: 'worktree' subagents running commands against main repo

    Resolved an issue where `isolation: 'worktree'` subagents could incorrectly execute Git-mutating commands against the main repository checkout. They will now correctly operate within their isolated worktrees.

  • Fix ultracode keyword opt-in firing on non-human-originated input

    The `ultracode` keyword opt-in will no longer trigger for non-human-originated input, such as webhook payloads and relayed PR comments. This ensures it only activates for direct user input.

  • Fix rendered text fragment leaking into crash telemetry

    Addressed a bug where a rendered text fragment could leak into crash telemetry when a UI component returned content outside a styled text element, improving telemetry accuracy.

  • Fix paste markers leaking into external editors

    Resolved an issue where paste markers were leaking into external editors opened from Claude Code, which could manifest as stray È/É characters around pasted text.

Read the original announcement →

https://github.com/anthropics/claude-code/releases/tag/v2.1.210

Related releases