Claude Code v2.1.141: New Environment Variables, Hooks Enhancement, and Bug Fixes
Anthropic shipped Claude Code v2.1.141 with a batch of quality-of-life improvements, new environment variables, and a long list of bug fixes. Here's what's relevant for developers running Claude Code in headless CI, multi-workspace setups, or with custom plugins.
New Environment Variables
CLAUDE_CODE_PLUGIN_PREFER_HTTPS — when set, GitHub plugin sources are cloned over HTTPS instead of SSH. Useful for environments without a configured SSH key.
ANTHROPIC_WORKSPACE_ID — scopes the minted token from workload identity federation to a specific workspace. Required when your federation rule covers multiple workspaces.
Hooks: Desktop Notifications Without a Terminal
Hooks now support a terminalSequence field in JSON output. This lets hooks emit desktop notifications, window title changes, and terminal bell sequences even when no controlling terminal is attached. Useful for long-running background agents in CI or daemon mode.
Agent & Session Improvements
- New
claude agents --cwd <path>flag scopes the session list to a specific directory. - Background agents launched via
/bgor←←now preserve the current permission mode instead of resetting to defaults. - Agents that finish work but leave a background shell running now appear under Completed, not Working.
- Feedback (
/feedback) can include recent sessions (last 24 hours or 7 days) for multi-session issues.
Rewind Menu: Summarize Up to Here
A new option Summarize up to here lets you compress earlier conversation context while keeping recent turns intact — helpful for long sessions without losing recent context.
Miscellaneous Changes
- Spinner now warms to amber after 10 seconds of thinking, so you know Claude hasn't frozen.
- Plugin menu navigation improved:
→/Tab cycles tabs,↑moves to tab strip, headers and search box are clickable in fullscreen. - Auto mode permission dialog now explains when a
permissions.askrule triggered it. - File-edit permission prompts now show the view diff in your IDE option again when an IDE is connected.
Bug Fixes (Highlights)
- Fixed background side-queries sending unavailable Haiku model ID on Bedrock/Vertex/Foundry/gateway — now falls back to the main-loop model.
- Fixed
claude daemon statusand/doctoron Windows when the daemon pipe key file is locked. - Fixed
/modelchanges in one session affecting autocompact threshold in other sessions. - Fixed markdown tables with cell wrapping falling back to key-value layout (regression in 2.1.136).
- Fixed cancelled prompts being dropped from Up-arrow history.
- Fixed
Ctrl+Cnot interrupting a running turn in vim INSERT/VISUAL mode. - Fixed
chat:submitkeybindings (meta+enter, ctrl+enter) not working when Enter is rebound tochat:newline. - Fixed
light-ansitheme using invisible white for diff context lines on light backgrounds. - Fixed error overlay dumping minified bundle source that hid the original error message.
📖 Read the full source: GitHub Claude-Code
👀 See Also
Cloken: A Chrome Extension That Shows Real-Time Claude Context Usage as a Percentage
Cloken is a free Chrome extension that displays your current Claude.ai chat context usage as a percentage — including messages, files, images, and system prompt.
Open Source Second Brain System Built on Claude Code for Task Management
An open source system called Kipi System uses Claude Code to track open threads, draft follow-ups, and manage tasks by pulling from calendar, email, CRM, and social feeds. It generates a daily HTML file with pre-written actions sorted by friction.
Single-call MCP pipeline reduces Claude Code token usage by 74%
A developer built a context engine MCP server that provides Claude Code with a dependency graph of codebases, reducing token usage by 65% initially. A new single-call pipeline further cuts tokens by 74% by eliminating multiple round trips and deduplicating results server-side.
Layerkit: AI Image Editor with Editable Layers Built with Claude Code
A developer built Layerkit, a browser-based AI image editor that generates scenes with editable layers to avoid constant re-prompting. The tool uses a multi-stage AI pipeline where one LLM plans composition, an image model generates the scene, and another LLM analyzes the actual image to place readable text.