Claude-Code v2.1.105 Release: Worktree Improvements, Plugin Monitors, and UI Fixes

New Features and Improvements
Claude-Code v2.1.105 introduces several functional enhancements and quality-of-life improvements for developers using the AI coding assistant.
The EnterWorktree tool now accepts a path parameter to switch into an existing worktree of the current repository, providing more flexibility in repository management workflows.
Plugin developers gain new capabilities with PreCompact hook support: hooks can now block compaction by exiting with code 2 or returning {"decision":"block"}. Background monitor support is added via a top-level monitors manifest key that auto-arms at session start or on skill invoke.
The command /proactive is now an alias for /loop, offering alternative command options for users.
Performance and Reliability Updates
API stream handling is improved: streams now abort after 5 minutes of no data and retry non-streaming instead of hanging indefinitely. Network error messages are enhanced to show a retry message immediately instead of displaying a silent spinner.
The WebFetch tool is improved to strip <style> and <script> contents from fetched pages, preventing CSS-heavy pages from exhausting the content budget before reaching actual text.
Stale agent worktree cleanup now removes worktrees whose PR was squash-merged instead of keeping them indefinitely. MCP large-output truncation prompts are enhanced to give format-specific recipes (e.g., jq for JSON, computed Read chunk sizes for text).
UI and Display Improvements
File write display is optimized: long single-line writes (e.g., minified JSON) are now truncated in the UI instead of paginating across many screens. The /doctor layout is improved with status icons, and users can press f to have Claude fix reported issues.
/config labels and descriptions are updated for clarity. Skill description handling is enhanced with the listing cap raised from 250 to 1,536 characters, and a startup warning is added when descriptions are truncated.
Bug Fixes
This release addresses over 30 issues, including:
- Images attached to queued messages (sent while Claude is working) being dropped
- Screen going blank when the prompt input wraps to a second line in long conversations
- Leading whitespace getting copied when selecting multi-line assistant responses in fullscreen mode
- Leading whitespace being trimmed from assistant messages, breaking ASCII art and indented diagrams
- Garbled bash output when commands print clickable file links (e.g., Python rich/loguru logging)
- Alt+enter not inserting a newline in terminals using ESC-prefix alt encoding, and Ctrl+J not inserting a newline (regression in 2.1.100)
- Duplicate "Creating worktree" text in EnterWorktree/ExitWorktree tool display
- Queued user prompts disappearing from focus mode
- One-shot scheduled tasks re-firing repeatedly when the file watcher missed the post-fire cleanup
- Inbound channel notifications being silently dropped after the first message for Team/Enterprise users
- Marketplace plugins with
package.jsonand lockfile not having dependencies installed automatically after install/update - Marketplace auto-update leaving the official marketplace in a broken state when a plugin process holds files open during the update
- Stdio MCP server emitting malformed (non-JSON) output hanging the session instead of failing fast with "Connection closed"
- MCP tools missing on the first turn of headless/remote-trigger sessions when MCP servers connect asynchronously
/modelpicker on AWS Bedrock in non-US regions persisting invalidus.*model IDs tosettings.jsonwhen inference profile discovery is still in-flight- 429 rate-limit errors showing a raw JSON dump instead of a clean message for API-key, Bedrock, and Vertex users
- Crash on resume when session contains malformed text blocks
/helpdropping the tab bar, Shortcuts heading, and footer at short terminal heights- Malformed keybinding entry values in
keybindings.jsonbeing silently loaded instead of rejected with a clear error CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFICin one project's settings permanently disabling usage metrics for all projects on the machine- Washed-out 16-color palette when using Ghostty, Kitty, Alacritty, WezTerm, foot, rio, or Contour over SSH/mosh
- Bash tool suggesting
acceptEditspermission mode when exiting plan mode would downgrade from a higher permission level
This release is for developers using Claude-Code who want improved worktree management, better plugin monitoring capabilities, and numerous bug fixes affecting UI display, terminal compatibility, and MCP server handling.
📖 Read the full source: GitHub Claude-Code
👀 See Also

Granite 4.1: IBM's 8B Dense Model Matches 32B MoE in Benchmarks
IBM's Granite 4.1 8B dense model matches or beats the previous 32B MoE model on ArenaHard, BFCL V3, GSM8K, and more, thanks to improved training data quality.

Claude Lacks Engineering Memory: On-Call Incident Reveals Missing Episodic Recall for Debugging Journeys
A developer spent 10 hours debugging a Kafka burst issue in a 1500-file monorepo, only to realize they had solved the exact same problem 4 months earlier — revealing that AI coding assistants like Claude lack episodic memory for past debugging journeys.

Open Source vs Frontier Models: Single-File Canvas Car Scene Benchmark
A developer tested 12 models including GPT-5.5, Claude Opus 4.7, and Qwen 3.6 Plus on a single-file HTML canvas car driving animation task, with results publicly compared.

Microsoft's BitNet Enables 100B Parameter LLM Inference on Single CPU
Microsoft's open-source BitNet project achieves 100B parameter LLM inference at 5-7 tokens/second on a single CPU, with the 2B parameter model using 0.4GB memory and 29ms latency while matching full-precision models on benchmarks.