v2.1.186
Claude CodeSummary
This release introduces new CLI commands for MCP server authentication (`login/logout`) and adds status filtering to the `/workflows` view. Numerous bugs have been resolved, particularly around session streaming stability after sleep, background agent interaction, and improved permission handling for subagents. Additionally, memory usage is improved via proactive index compaction.
New Features
- Added `claude mcp login <name>` and `claude mcp logout <name>` commands to authenticate MCP servers directly from the CLI, supporting stdin redirect for SSH sessions.
- Added status filtering (press `f`) to the `/workflows` agent detail view for easier navigation.
- Added a "Skills" section to the `/plugin` Installed tab for better organization.
- Added `teammateMode: "iterm2"` setting to explicitly configure iTerm2 integration.
- Added "Claude Platform on AWS - refresh credentials" option to `/login` when `awsAuthRefresh` is configured.
- Enabled automatic Claude response to bash command output when using `!` prefix; this behavior can be disabled via settings.
- Changed background subagents to surface permission prompts in the main session instead of auto-denying, allowing users to explicitly grant tool access.
- Changed `/review <pr>` to utilize the same review engine as `/code-review medium`.
Bug Fixes
- Fixed streaming requests failing with "Content block not found" or JSON parse errors after the machine wakes from sleep.
- Fixed subagent transcript scroll position incorrectly bleeding into the main transcript upon subagent exit.
- Fixed background task previews briefly flashing raw tool names before the agent's plan loaded.
- Fixed Chrome tab-group isolation not applying correctly when the in-product permissions gate is disabled for concurrent CLI sessions.
- Fixed background session recaps being duplicated; the agent's own end-of-turn summary is now correctly shown as the recap line.
- Fixed opening a background session from `claude agents` leaving the previous screen content visible behind the new session.
- Fixed `Agent(type)` deny rules and `Agent(x,y)` allowed-types restrictions not being enforced during named subagent spawns.
- Fixed Esc and Ctrl+C not responding while background agents were still running after the main turn concluded.
- Fixed misaligned option numbers in permission prompts when the option text wrapped to a new line.
- Fixed pressing `x` on a finished subagent in the agent panel failing to dismiss it.
- Fixed a misleading "MCP server disconnected" notice appearing for intentionally retired tools when resuming older sessions.
- Fixed the `/plugin` Installed view showing a "more above" indicator when already scrolled to the top.
- Fixed `~~strikethrough~~ ` rendering as literal tildes instead of strikethrough formatting in assistant messages.
- Fixed `--tools` flag allowing feature-gated tools to load prematurely on a cold first launch.
- Fixed background job status in `claude agents` showing a stale "needs input" message after a reply was sent.
- Fixed a dark-theme flash occurring when opening a background session from `claude agents` in a light terminal theme.
- Fixed mouse-selected text remaining highlighted after being deleted within `claude agents`.
- Fixed session cost not displaying for usage-based Enterprise and Team subscribers.
- Fixed agent teams: teammates spawned via tmux/pane backends no longer inherit the leader's `--effort` level incorrectly.
- Fixed Workflow `agent({schema})` subagents looping indefinitely on repeated schema validation failures instead of aborting after 5 attempts.
Improvements
- Improved `claude mcp get` and `claude mcp remove` commands to suggest the closest configured server name upon a typo and truncate excessively long server lists.
- Improved memory management by reminding the agent to compact its `MEMORY.md` index when it approaches the size limit.
- Improved skill frontmatter parsing: `display-name`, `default-enabled`, `fallback`, and `metadata.*` keys now correctly accept kebab-case, snake_case, and camelCase.
- Improved handling of malformed `SKILL.md` YAML frontmatter by loading the skill body with empty metadata instead of failing silently.
- Changed the maximum retry count for `CLAUDE_CODE_MAX_RETRIES` to 15; users are advised to use `CLAUDE_CODE_RETRY_WATCHDOG` for unattended sessions.