Claude Code v2.1.186: MCP CLI Auth, Bash Auto-Reply

Claude Code v2.1.186 ships with a batch of practical improvements: MCP server authentication from the CLI, automatic responses to bash commands, and over 20 bug fixes spanning session persistence, subagent management, and UI polish.

CLI Authentication for MCP Servers

Two new commands let you authenticate MCP servers without dropping into the interactive /mcp menu:

claude mcp login <name>
claude mcp logout <name>

For SSH sessions, pass --no-browser and redirect stdin to complete the flow headlessly.

Bash Commands Now Auto-Reply

By default, ! bash commands now trigger Claude to automatically respond to the output—no more silent execution. To keep the previous behavior (bash output only added to context without a response), set "respondToBashCommands": false in settings.json.

Workflows & Subagents

Press f in the /workflows agent detail view to filter by status.
Background subagents now surface permission prompts in the main session instead of auto-denying. The dialog shows which agent is asking; Esc denies just that tool.
Agent(type) deny rules and Agent(x,y) allowed-types restrictions now actually enforce for named subagent spawns.
Workflow agent({schema}) subagents will abort after 5 repeated schema validation failures instead of looping forever.
Teammates spawned via tmux/pane backends now inherit the leader's --effort level.

Memory & Skills

The agent is now reminded to compact its MEMORY.md index when nearing the size limit.
Skill frontmatter keys (display-name, default-enabled, fallback, metadata.*) now accept kebab-case, snake_case, and camelCase.
Malformed SKILL.md YAML frontmatter loads the skill body with empty metadata instead of failing silently.

Settings & Integrations

Added teammateMode: "iterm2" setting; warns when auto mode cannot find the it2 CLI.
Added "Claude Platform on AWS - refresh credentials" option to /login when awsAuthRefresh is configured.
CLAUDE_CODE_MAX_RETRIES now caps at 15; for unattended sessions, use CLAUDE_CODE_RETRY_WATCHDOG.
/review <pr> now uses the same review engine as /code-review medium.

Bug Fixes (20+)

Notable fixes include streaming resuming after sleep (no more "Content block not found" or JSON parse errors), Chrome tab-group isolation applying correctly when the permissions gate is off for concurrent CLI sessions, background session recaps no longer duplicated, and ~~strikethrough~~ actually rendering as strikethrough instead of literal tildes. Also fixed: ESC/Ctrl+C responsiveness while background agents run, session cost now showing for usage-based Enterprise and Team subscribers, and a dark-theme flash when opening a background session from claude agents in a light terminal.

For the full changelog, see the source.

📖 Read the full source: GitHub Claude-Code