Claude Code Weekly Updates (2.1.43 - 2.1.50)

Posted on February 22, 2026

title

Here are the Claude Code updates from February 16-22, 2026 that I’m interested in.

TL;DR

version: 2.1.43 -> 2.1.50
changes: 142
Undocumented in release notes
- Context usage percentage shown during Plan approval
Key topics
- Claude Sonnet 4.6
- Full git worktree support (--worktree flag, subagent isolation, hooks)

v2.1.43 & v2.1.44 (5 changes)

2.1.43~2.1.44

v2.1.43 and v2.1.44 were mostly minor fixes. Nothing particularly noteworthy here.

v2.1.45 & v2.1.46 (17 changes)

2.1.45~2.1.46

Claude Sonnet 4.6 Support

Added support for Claude Sonnet 4.6

Claude Sonnet 4.6 has been added, bringing across-the-board upgrades in:

Coding
Computer use
Long-context reasoning
Agentic planning
Knowledge work
Design

Some early-access developers have rated it as outperforming Opus 4.5. Personally, I find it well-suited for documentation tasks and tool interactions. I’ve been using Sonnet 4.6 extensively with Claude Cowork lately.

Introducing Sonnet 4.6

Claude Sonnet 4.6 is a full upgrade of the model’s skills across coding, computer use, long-reasoning, agent planning, knowledge work, and design.

www.anthropic.com

Customizable Spinner Tips

Added spinnerTipsOverride setting to customize spinner tips. Configure tips with an array of custom tip strings, and optionally set excludeDefault: true to show only your custom tips instead of the built-in ones

You can now customize the tips shown in the spinner via the spinnerTipsOverride setting. Set your own array of tips, and optionally use excludeDefault: true to hide the defaults.

{
  "spinnarTipsOverride": {
    "excludeDefault": "true", // "false" default
    "tips": [
      "claude code charcter's name is Clawd",
      "Tasks was general-purpose subagent"
    ]
  }
}

As a side note, you can also customize spinnerVerbs since v2.1.23:

{
  "spinnerVerbs": {
    "mode": "replace", // or "append"
    "verbs": [
      "alpha",
      "bravo",
      "charlie",
      "delta",
    ]
  }
}

v2.1.47 (68 changes)

v2.1.47 key changes

`last_assistant_message` Added to Stop/SubagentStop Hooks

Added last_assistant_message field to Stop and SubagentStop hook inputs, providing the final assistant response text so hooks can access it without parsing transcript files.

The last_assistant_message field has been added to Stop and SubagentStop hooks. This lets hooks access the final response text directly, eliminating the need to parse transcript files.

Significant Memory Usage Improvements

Improved memory usage in long-running sessions by releasing API stream buffers, agent context, and skill state after use Improved memory usage by trimming agent task message history after tasks complete Improved memory usage during long agent sessions by eliminating O(n²) message accumulation in progress updates

Memory usage in long-running sessions has been improved from multiple angles. There have been frequent reports of Claude Code feeling sluggish recently, so this likely addresses those concerns.

Improvements include releasing API stream buffers, freeing agent context, and eliminating O(n²) message accumulation in progress updates.

Ctrl+F to Kill Background Agents

Use ctrl+f to kill all background agents instead of double-pressing ESC. Background agents now continue running when you press ESC to cancel the main thread, giving you more control over agent lifecycle.

Killing background agents has been moved to Ctrl+F. ESC now only cancels the main thread, allowing background agents to keep running.

Bulk Windows Fixes

This release includes a large number of Windows-related fixes:

Terminal rendering bugs (\r\n line count display issues)
Bash tool output being discarded on MSYS2/Cygwin
CWD tracking temp files accumulating indefinitely
Image paste support on WSL2 (BMP format)
Layout issues caused by CJK full-width characters
Right Alt key escape sequence residue

v2.1.49 (27 changes)

v2.1.49 key changes

`--worktree` Flag

Added --worktree (-w) flag to start Claude in an isolated git worktree

The --worktree (-w) flag lets you launch Claude Code in an isolated git worktree, so you can make experimental changes without touching your main working tree.

Two new startup flags were added in v2.1.49:

-w, —worktree [name]: Create a new git worktree for this session (optionally specify a name)
—tmux: Create a tmux session for the worktree (requires —worktree). Uses iTerm2 native panes when available; use —tmux=classic for traditional tmux.

The —worktree flag generates a worktree directory under ./.claude/worktree. If you don’t specify a [name], a random human-readable name is assigned.

The —tmux flag can only be used alongside —worktree and launches a tmux session.

Subagent Worktree Isolation

Subagents support isolation: "worktree" for working in a temporary git worktree

Setting isolation: "worktree" on subagents lets them work in a temporary git worktree. This means Agent Teams teammates can work in parallel without stepping on each other’s changes.

---
name: safe-researcher
description: Research agent with restricted capabilities
tools: Read, Grep, Glob, Bash
isolation: worktree
---

ConfigChange Hook Event

Added ConfigChange hook event that fires when configuration files change during a session, enabling enterprise security auditing and optional blocking of settings changes.

A new hook event fires when config files change during a session. This enables enterprise security auditing and the ability to block settings changes.

Sonnet 4.5 [1M] Gradual Transition

Sonnet 4.5 with 1M context is being removed from the Max plan in favor of our frontier Sonnet 4.6 model, which now has 1M context. Please switch in /model.

Sonnet 4.5 [1M] on the Max plan is being replaced by Sonnet 4.6 [1M].

v2.1.50 (25 changes)

v2.1.50 key changes

WorktreeCreate/WorktreeRemove Hooks

Added WorktreeCreate and WorktreeRemove hook events, enabling custom VCS setup and teardown when agent worktree isolation creates or removes worktrees.

Hook events for worktree creation and removal have been added. Combined with the --worktree flag from v2.1.49, you now have hook-based control over the entire worktree lifecycle.

Declarative Worktree Isolation in Agent Definitions

Added support for isolation: worktree in agent definitions, allowing agents to declaratively run in isolated git worktrees.

You can now specify isolation: worktree directly in agent definition files (.claude/agents/*.md).

`claude agents` CLI Command

Added claude agents CLI command to list all configured agents

A new CLI command to list all configured agents. You can list them with claude agents.

claude agents

Memory Leak Fixes

Fixed memory leak in agent teams where completed teammate tasks were never garbage collected from session state Fixed a memory leak where LSP diagnostic data was never cleaned up after delivery, causing unbounded memory growth in long sessions Fixed a memory leak where completed task output was not freed from memory, reducing memory usage in long sessions with many tasks

v2.1.50 includes a wave of memory leak fixes covering Agent Teams tasks, LSP diagnostic data, CircularBuffer, shell command execution, TaskOutput, and file history snapshots. Memory consumption in long-running sessions should be significantly improved.

Undocumented Feature

Context Usage Percentage on Plan Approval

Context usage shown during Plan approval

When approving a plan in Plan mode, the current context usage is now displayed as a percentage. This wasn’t mentioned in the release notes.

Previously, you could see the context window size via the status line, but it wasn’t visible during plan approval. Since many users want to decide whether to clear context based on remaining window size, this is a welcome addition.