v2.0.0 → v2.1.220 · 2025.9.29 – 2026.7.25
Claude Code Opus 기본 모델 교체 — 1M 토큰 컨텍스트 · fast mode 지원.
/fast 토글; Opus 5 vs Sonnet 5 선택 기준컨텍스트 한도 초과 시 명시적 경고 — /compact 실패는 에러로 표시.
코드 리뷰가 백그라운드에서 실행 — 스택된 슬래시 명령도 리뷰 대상 유지.
Claude가 자동으로 실행하지 않음 — 원할 때 /verify·/code-review 직접 호출.
/code-review, 머지 전 /verify 직접 실행오래 걸리는 도구 호출이 침묵하지 않고 주기적으로 진행 상황을 전송.
vim 모드 삽입 상태에서 두 키 시퀀스를 원하는 키로 리맵.
vimInsertModeRemaps에 시퀀스→키 맵핑 추가Bedrock·Vertex AI·Foundry에서 Auto mode 기본 활성 — disableAutoMode 설정으로 비활성화.
CLAUDE_CODE_ENABLE_AUTO_MODE 환경변수 불필요 — 클라우드 플랫폼에서 기본 활성disableAutoModeClaude Code 기본 모델 교체 — 네이티브 1M 토큰 컨텍스트 탑재.
수출통제 중단 후 Mythos급 Fable 5 재배포.
"워크플로 만들어줘" → 백그라운드서 수십~수백 에이전트로 오케스트레이션.
v2.1.219부터 중첩 spawn 기본 활성(최대 depth 3) — CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH=1로 비활성화.
CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH=1로 기존 동작 복원; 동시성 상한 CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS기본 모델 과부하·불가 시 최대 3개를 순서대로 폴백. --fallback-model 로도.
CLAUDE.md·plugins·skills·hooks·MCP 모두 비활성으로 기동(디버그용). CLAUDE_CODE_SAFE_MODE 도.
세션 도중 작업 폴더 변경 — 프롬프트 캐시 안 깨고.
프롬프트에서 직접 설정 오버라이드 — interactive·-p·Remote Control 모두 지원.
/config thinking=false 활용! 명령 실행 후 Claude가 출력을 자동으로 분석·응답.
respondToBashCommands: falsebg 에이전트 권한 요청이 자동 거부 대신 메인 세션에 표시 — Esc로 해당 도구만 거부.
.env·~/.aws/credentials 등 자격증명 파일을 샌드박스 도구로부터 보호하는 설정.
/clear 실행 직전 상태로 대화를 복원해 재개.
! 명령 입력 시 파일 경로를 탭으로 자동완성.
워크트리 코드 작업 완료 시 커밋·푸시·드래프트 PR까지 자동 처리.
입력 필요·완료 이벤트에 Notification hook 발화 — 외부 알림 자동화 가능.
/skill-a /skill-b task 형식으로 최대 5개 스킬을 순서대로 한 번에 로드.
환경·설정 전체 진단 후 자동 수정 — /checkup으로도 사용 가능.
-p 모드 stream-json 파이프라인에서 서브에이전트 출력까지 캡처하는 플래그·환경변수.
CLAUDE_CODE_FORWARD_SUBAGENT_TEXT 환경변수도 지원worktree에서 승인한 권한이 저장소 루트에 저장 — 세션·워크트리 전체에 자동 적용.
/fork가 현재 대화를 새 bg 세션으로 복사; 인세션 위임 실행은 /subtask로 명칭 변경.
effort 선택, --comment로 인라인 게시
skills·subagents·plugins·MCP별 토큰 비용
화살표·j/k·Space로 diff 탐색
유휴·업데이트 시에도 유지
live 세션을 JSON으로 출력
v/V로 선택 + 오퍼레이터
타이핑으로 긴 스킬 목록 검색
비대화형으로 ultrareview 실행
자동설치된 의존성 제거
theme·editor 등 settings.json에
@로 이름 지은 서브에이전트 호출
Bash(git *) 같은 규칙으로 필터
headless에서 도구 앞 일시정지
분류기 거부 시 retry 가능
코드베이스 분해 → worktree별 실행
코드블록 선택해 클립보드로
에이전트가 worktree에서 실행
URL로 JSON POST·수신
외부 빌드에 로컬 환경 서빙
컨텍스트별·코드 시퀀스 지정
GitHub PR 번호로 resume
커스텀 명령에서 개별 인자 접근
현재 브랜치 PR 상태 표시
다른 디렉토리 CLAUDE.md 로드
go-to-definition 등 LSP 기능
브라우저 연동 (Beta)
Kitty·Alacritty·Zed·Warp 지원
[Image #N] 클릭으로 뷰어
Claude Opus 4.5 추가
Azure AI Foundry 연동
도구 권한 자동 승인/거부
서브에이전트 시작 시 트리거
재사용 가능한 Claude Skills
Claude이 객관식으로 물어봄
모델 선택기에 Haiku 4.5
allowUnsandboxedCommands 등
stop 시 프롬프트 기반 훅
대화+코드 한 번에 롤백
플랜 한도 확인
=1 설정
v2.1.219
/verify·/code-review 직접 호출
v2.1.215
CLAUDE_CODE_MAX_WEB_SEARCHES_PER_SESSION
v212
/compact 실패도 에러로 표시 (무응답 없음)
v2.1.216
/checkup 별칭
v205
.claude/settings.json의 pluginConfigs 무시 — user·--settings·managed만 인정
v207
CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS로 임계값 조정
v212
OTEL_LOG_ASSISTANT_RESPONSES=1로 활성화
v193
${user_config.*} 금지 — exec form 또는 $CLAUDE_PLUGIN_OPTION_KEY 사용
v207
CLAUDE_CODE_FORWARD_SUBAGENT_TEXT도 지원
v211
if:의 dir/**은 <cwd>/dir만 매칭 — 재귀 매칭은 **/dir/** 사용
v214
register_repo_root로 새 작업 디렉터리 등록 시 발화
v2.1.219
CLAUDE_CODE_ENABLE_AUTO_MODE 불필요; disableAutoMode로 끄기
v207
claude-opus-5) 추가, 이제 기본 Opus 모델 — 1M 컨텍스트, fast mode $10/$50/Mtok
sandbox.network.strictAllowlist 설정 추가
/add-dir 또는 SDK register_repo_root 컨트롤 요청으로 새 작업 디렉터리 등록 후 발동하는 DirectoryAdded 훅 추가
mcp_server_errors 추가 — 설정 유효성 검사에서 제외된 --mcp-config 항목 나열; 터미널 실행 시 시작 경고 출력
workflowSizeGuideline 설정 키 추가; 설정 시 /config 행 숨김
--forward-subagent-text 설정 시 depth-2+ 서브에이전트가 스폰한 Agent tool_use id로 키잉되어 출력
claude mcp list·/mcp에 HTTP 상태·오류 텍스트 추가, 앞뒤 숨겨진 공백이 있는 MCP 설정값 경고 추가
/config의 Dynamic workflow size로 다른 크기·무제한 선택 가능
${VAR} 항목이 설정 파일 env 대신 시작 환경·managed-settings env에서 해석되도록 변경
/model 피커가 가장 최신 모델 이름만 강조하도록 변경 — 임의 목록이 아닌 새 릴리즈를 명확히 표시
/config에서 변경 안내 포함
/fast가 이제 Opus 5·Opus 4.8에 적용
claude --teleport to show which repo your current checkout points at when it doesn't match the session's repo/code-review이 백그라운드 서브에이전트로 실행되도록 변경 — 리뷰 작업이 대화를 채우지 않고 스택된 슬래시 커맨드를 리뷰 대상으로 유지
--ax-screen-reader 모드에서 단어·줄 삭제(Option+Delete·Ctrl+W·Cmd+Backspace·Ctrl+U·Ctrl+K) 시 삭제된 텍스트를 스크린 리더로 알림
claude mcp list와 /mcp에 HTTP 상태·오류 텍스트 추가, 앞뒤에 숨겨진 공백이 있는 MCP 설정값 경고 추가
/deep-research가 수동으로 호출할 때만 시작하도록 변경 — Claude가 자체적으로 실행하지 않음
/config model=<x> 또는 Remote Control로 모델 전환 시 fast mode 변경 알림 추가
:을 포함한 에이전트 이름을 거부하도록 변경
context: fork가 있는 스킬이 기본적으로 백그라운드에서 실행되도록 변경; 스킬별로 background: false로 비활성화 가능
true/false 외에 yes/no/on/off/1/0(대소문자 무관) 허용
/ultrareview error feedback so Claude can correct an invalid argument instead of retrying it unchanged&, and suspicious-Windows-path checks no longer open permission dialogs; the auto-mode classifier adjudicates them instead/ultrareview failing on descriptive arguments like "review my auth changes" — they now run a review of your current branch with the text applied as a note to the findings:heart:를 입력하면 ❤️ 삽입, :hea는 제안 표시 — emojiCompletionEnabled 설정으로 비활성화
CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS로 재정의) 추가
CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH 설정
FORCE_HYPERLINK=0 to opt outOTEL_EXPORTER_OTLP_ENDPOINT not governing all signals — lower-scope signal-specific overrides no longer redirect telemetry away from the managed endpointsandbox.filesystem.disabled 설정 추가
/context가 대화가 컨텍스트 윈도를 초과할 때 명시적 경고 표시; 실패한 /compact는 오류로 표시
/rewind이 추적 경로의 심링크·하드링크를 통해 파일을 복원하거나 삭제하지 않으며 건너뛴 경로 수를 보고
/mcp와 /install-github-app이 에이전트 뷰에 "needs input" 요청 저장
/fork confirmation to one line with the new session's name, claude attach id, and a note when the copy shares your checkoutgit and gh command arguments in the PowerShell tool/ultrareview diff-too-large error to show configured limits, measured diff size, and largest contributing files/code-review ultra empty-diff message to name the exact base ref and suggest passing an explicit base&& lists or negations/verify와 /code-review 스킬을 더 이상 자율 실행하지 않음; 원할 때 직접 /verify 또는 /code-review로 호출
modified 타임스탬프 추가
message.uuid·client_request_id·tool_source 속성 추가
CLAUDE_CODE_OTEL_CONTENT_MAX_LENGTH 추가
subagentStatusLine 페이로드에 reasoning effort 추가
--url·--connection·--identity 및 Podman 원격 모드)를 포함한 docker 명령(Podman docker shim 포함)에 기존에 없던 권한 프롬프트 추가
dir/ 훅 if: 조건이 <cwd>/dir에만 매칭되도록 변경; 임의 깊이 매칭은 /dir/**로 작성. deny/ask 권한 규칙은 임의 깊이 매칭 유지.
-m/--magic-file 또는 -f/--files-from을 사용하는 file 명령이 읽기 전용으로 자동 허용되지 않고 권한 요청하도록 변경
"resume" 대신 "fork"로 소스를 보고하도록 변경
claude rc workspace-trust error in the home directory to say trust there is never saved and to suggest running from a project directory[[ ]] comparisons as inert text — these commands now prompt for approvalhelp and man commands that could run unsafe options, command substitutions, or backslash paths/fork이 이제 작업을 유지한 채 대화를 새 백그라운드 세션(claude agents의 별도 행)으로 복사; 기존에 실행하던 인세션 서브에이전트는 이제 /subtask
claude auto-mode reset 추가, 확인 프롬프트 포함(--yes로 건너뜀)
CLAUDE_CODE_MAX_WEB_SEARCHES_PER_SESSION으로 조정) 추가
CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION으로 재정의) 추가; /clear로 예산 초기화
CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS
/resume 입력 시 목록에서 삭제된 세션 포함 과거 세션 피커가 열리고 선택한 세션이 백그라운드로 재개
/fork 복사본을 입력한 프롬프트로 이름 지정해 에이전트 뷰에서 식별 가능하도록 변경
/btw가 가장 최근 교환의 사이드 질문 패널을 다시 열어 이전 답변 탐색 가능하도록 변경
← 푸터 힌트가 잠시 N done으로 반짝이도록 변경
mode 파라미터 폐기(이제 무시됨); 서브에이전트가 기본적으로 부모 세션의 권한 모드 상속
forceLoginMethod가 터미널뿐 아니라 VS Code 익스텐션·SDK·setup-token·install-github-app 로그인에도 적용되도록 변경
set_model 컨트롤 요청을 턴 중간에 적용; 다음 모델 라운드트립에서 다음 턴을 기다리지 않고 새 모델 사용
claude agents --json: 샌드박스·MCP 입력·관리형 설정 프롬프트 대기 세션이 "Working" 대신 "Needs input"으로 표시
SendMessage bodies are no longer duplicated into replayed history and tool resultstouch, rm) without a permission prompt or SDK canUseTool callback--forward-subagent-text 플래그와 CLAUDE_CODE_FORWARD_SUBAGENT_TEXT 환경 변수 추가
/clear 후 CCR 웹 fetch/search 프록시 정상 작동 유지
1e6·64_000 같은 지수 표기 및 숫자 구분자 형식 허용
/usage-credits가 조직 관리자에게 요청 전송 전 확인 요청하도록 변경
s·S(문자/줄 치환)가 NORMAL 모드에서 vim 동작에 맞게 작동하도록 변경
save_to_disk가 이미지를 디스크에 저장하고 경로를 반환(기존에는 동작 없음)
ask decision for unsandboxed Bash — a hook ask now floors the decision at a promptANTHROPIC_AUTH_TOKEN + ANTHROPIC_BASE_URL) coming back "Not logged in" after the daemon respawns themWrite(path)·NotebookEdit(path)·Glob(path) 권한 규칙에 시작 시 경고 추가 — 대신 Edit(path)·Read(path) 사용
isolation: 'worktree' subagents being able to run git-mutating commands against the main repo checkout instead of their own isolated worktree.claude/* symlinks not being reconciled into the sandbox deny-write listclaude --ax-screen-reader 실행, CLAUDE_AX_SCREEN_READER=1 설정, 또는 settings에 "axScreenReader": true 추가
jj 같은 두 키 삽입 시퀀스를 Escape에 매핑하는 vimInsertModeRemaps 설정 추가
CLAUDE_CODE_PROCESS_WRAPPER 추가 — 에이전트 뷰·백그라운드 서비스가 모든 Claude Code 자체 spawn을 지정 래퍼 실행 파일을 통해 실행(기업 런처 준수)
/tasks에 유지
$(…)/백틱/<(…) 포함 명령의 위험 삭제(예: rm -rf ~)가 --dangerously-skip-permissions·auto 모드에서도 일반 형식처럼 프롬프트
/install-github-app과 /mcp 설정 메뉴가 백그라운드 세션에서 더 이상 열리지 않음
/mcp에 "not configured"로 표시
/usage가 마지막 사용량 바와 "as of" 메모 표시
CLAUDE_CODE_ENABLE_AUTO_MODE opt-in 없이 auto 모드 사용 가능; settings에서 disableAutoMode로 비활성화
.claude/settings.local.json에서 autoMode를 더 이상 읽지 않도록 변경; 대신 ~/.claude/settings.json 사용
${user_config.*}가 이제 거부됨(셸 인젝션 수정). 훅: exec 형식(args 배열) 또는 $CLAUDE_PLUGIN_OPTION_<KEY> 사용; 모니터·headersHelper: 스크립트 내에서 값 직접 읽기(설정 파일 또는 서버의 env 블록)
pluginConfigs)이 프로젝트 수준 .claude/settings.json에서 더 이상 읽히지 않음; 사용자·--settings·관리형 설정만 적용
[Pasted text #N] placeholder instead of adding a second onewaiting 3m) instead of the same timestamp twiceclaude -p, the SDK) being permanently recorded as consented without ever showing the security consent dialogcd prompting for permission when the only output redirect was to /dev/nullcredential_process): the 60-second stall guard now fires instead of waiting forever./add-dir 동작과 동일하게 /cd에 디렉터리 경로 자동완성 추가
/doctor 검사 추가 — 저장된 CLAUDE.md 파일 경량화
/commit-push-pr가 origin 외에도 저장소에 설정된 push 리모트(remote.pushDefault, 또는 유일한 리모트)로의 git push를 자동 허용
/login이 Anthropic 운영 퍼블릭 게이트웨이 엔드포인트 지원
EnterWorktree가 프로젝트 .claude/worktrees/ 외부 git worktree 진입 전 확인 요청
awsCredentialExport 헬퍼 사용 시 수 분간 시작 중단 수정
/code-review findings quality on claude-opus-4-8 across all effort levels--permission-prompt-tool pointing at an MCP server crashing with "MCP tool not found" on cold start before the server finishes connecting/model picker when an entitlement or allowlist restriction drops the row they were positioned against/doctor가 문제를 진단·수정하는 전체 설정 검사 도구로 강화; /checkup은 별칭
rm -rf on a variable it can't resolve from contextclaude agentsroots/list에 추가 — 변경 시 notifications/roots/list_changed 전송
claude agents 뷰에서 섹션(Needs input / Working / Completed)이 항상 표시되도록 변경
/doctor·/status로 이동
claude agents 푸터의 중복 탐색 힌트 제거
/config에 "Dynamic workflow size" 설정 추가 — 동적 워크플로 규모 제어(small/medium/large 에이전트 수, 권고 기준이며 강제 상한 아님)
workflow.run_id·workflow.name OpenTelemetry 속성 추가 — OTel 데이터로 워크플로 실행 활동 재구성 가능
/review <pr>를 빠른 단일 패스 리뷰로 복원 — 멀티 에이전트 리뷰는 /code-review <level> <pr#> 사용
/workflows agent list layout: wider titles, a dedicated time column, shorter model names, and no per-row tool-call countsurl but no type, suggesting "type": "http" instead of the misleading "command: expected string"claude auth login and claude mcp login --no-browser not being reliably clickable when it wraps over SSH — it is now emitted as a single hyperlink/remote-control sessions showing the wrong permission mode in the mobile and web appsAskUserQuestion 대화상자가 기본적으로 자동 계속하지 않도록 변경; /config에서 유휴 타임아웃 opt-in
--help·VS Code·JetBrains에서 "default" 권한 모드를 "Manual"로 변경; default와 함께 --permission-mode manual·"defaultMode": "manual" 허용
Header: value. lines/skill-a /skill-b do XYZ 같은 슬래시 스킬 중첩 호출 시 첫 번째만이 아닌 앞쪽 스킬 전부(최대 5개) 로드
CLAUDE_CODE_RETRY_WATCHDOG이 비용량 일시 오류의 기본 재시도 횟수를 300으로 올리고 CLAUDE_CODE_MAX_RETRIES의 15 상한 해제
claude agents 세션 행의 풀 리퀘스트 링크가 불필요한 "PR" 레이블 없이 #N만 표시
claude --dangerously-skip-permissions daemon <subcommand> being treated as a chat prompt instead of running the subcommandclaude agents에 백그라운드 에이전트 알림 추가 — 입력 대기·완료 세션이 Notification 훅 발동 (agent_needs_input / agent_completed)
/dataviz 스킬 추가
claude agents에서 시작된 백그라운드 에이전트가 worktree 코드 작업 완료 시 커밋·푸시·draft PR 자동 오픈(질문 없이)
claude agents 뷰에서 /login이 "사용 불가" 대신 로그인 대화상자 오픈
/agents 위저드 제거 — Claude에게 서브에이전트 생성·관리를 요청하거나 .claude/agents/를 직접 편집
awsAuthRefresh now runs automatically/model에 "Org default"(또는 "Role default")로 표시
claude mcp list/get가 커밋된 .claude/settings.json으로 자체 승인된 .mcp.json 서버를 더 이상 spawn하지 않음 — 신뢰 안 된 워크스페이스는 ⏸ Pending approval 표시
ANTHROPIC_BASE_URL이 비Anthropic 호스트를 가리키면 Remote Control 비활성화 — CLAUDE_CODE_USE_BEDROCK/_VERTEX/_FOUNDRY 기존 동작과 일치
← 한 번으로 변경 — 백그라운드 세션 동작에 맞춤
/code-review workflow: merged five cleanup finders into one, cutting token usage by roughly 25%CLAUDE_ENABLE_STREAM_WATCHDOG=0 to disable.claude agents --dangerously-skip-permissions silently falling back to auto mode instead of showing the bypass disclaimer and applying bypass mode to spawned agentsscopes_supported catalog when no scope is specified, causing invalid_scope failures on GitLab self-hosted and other enterprise IdPsCLAUDE_CODE_DISABLE_MOUSE_CLICKS 추가
claude agents completed list to fill available vertical space; on short terminals the header compacts so live sessions stay visibleautoMode.classifyAllShell 설정 추가
/permissions 최근 거부 내역에 auto-mode 거부 사유 추가
claude_code.assistant_response OpenTelemetry 로그 이벤트 추가. OTEL_LOG_ASSISTANT_RESPONSES=1 없으면 redact; 미설정 시 OTEL_LOG_USER_PROMPTS 따름(프롬프트 로깅 환경은 업그레이드 후 응답도 수신됨) — 프롬프트만 유지하려면 OTEL_LOG_ASSISTANT_RESPONSES=0 설정
!)에 실시간 파일 경로 자동완성 추가
/mcp를 안내하는 시작 알림 추가
CLAUDE_CODE_DISABLE_BG_SHELL_PRESSURE_REAP=1)
headersHelper auth: the helper now re-runs and reconnects automatically when a tool call returns 401/403renames maps are now followed automatically, updating your settings to the new name/add-dir message when the directory is already a working directory/clear 실행 전 시점으로 대화를 재개하는 /rewind 지원 추가
tools/list, prompts/list, resources/list) now retries transient network errors with short backoff/) to hint how to reach slash commands/voice showing a generic "not available" message when disabled by an organization's policy — it now explains the restriction/permissions Recently-denied tab: approving a denial now persists on close instead of being silently discardedforceRemoteSettingsRefresh now takes effect when set via MDM or file policy, and the fetch sends Cache-Control: no-cache to prevent proxies from serving stale responsessandbox.credentials 설정 추가
--model·/model·ANTHROPIC_MODEL에 적용, 제한 모델 선택 시 "restricted by your organization's settings" 메시지 표시
/model·/config 등)에서 마우스 클릭 지원 추가
/install-github-app: GitHub Actions workflow setup is now optional — you can install just the GitHub App and skip the workflow/secret steps/btw with ←/→ arrow navigation to step through earlier answers/plugin to surface plugins you haven't used recently so you can clean them up.git/worktrees/ entries from killed agents are now cleaned up automatically/mcp 메뉴 없이 MCP 서버 인증하는 claude mcp login <name>·claude mcp logout <name> 추가, SSH 완료용 --no-browser stdin 리다이렉트 지원
/workflows 에이전트 상세 뷰에 상태 필터링(f 키) 추가
/plugin 설치됨 탭에 "Skills" 섹션 추가
it2 CLI를 찾지 못할 때 경고를 표시하는 teammateMode: "iterm2" 설정 추가
awsAuthRefresh 설정 시 /login에 "Claude Platform on AWS - refresh credentials" 옵션 추가
! bash 명령 출력에 Claude가 자동 응답; 기존 컨텍스트 전용 동작 유지는 settings.json에 "respondToBashCommands": false 설정
CLAUDE_CODE_MAX_RETRIES 최대 15로 제한; 무인 세션에는 CLAUDE_CODE_RETRY_WATCHDOG 사용
/review <pr>가 /code-review medium과 동일한 리뷰 엔진 사용으로 변경
claude mcp get and claude mcp remove to suggest the closest configured server name on a typo and truncate long server listsMEMORY.md index when nearing the size limitdisplay-name, default-enabled, fallback, and metadata.* keys now accept kebab-case, snake_case, and camelCaseSKILL.md YAML frontmatter handling: loads the skill body with empty metadata instead of failing silentlyAgent(type) deny rules and Agent(x,y) allowed-types restrictions not being enforced for named subagent spawns-p)에서 stderr 출력, agent frontmatter 설정 모델도 포함
attribution.sessionUrl 설정 추가
/config key=value에 쓸 수 있는 모든 단축 키 목록을 보여주는 /config --help 추가
/config 토글 동작 변경 — Enter·Space 모두 선택 설정 변경, Esc는 이제 되돌리기 대신 저장 후 닫기
/doctor, 또는 --debug 사용
git reset --hard, git checkout -- ., git clean -fd, git stash drop) are now blocked when you didn't ask to discard local work, git commit --amend is blocked when the commit wasn't made by the agent this session, and terraform destroy/pulumi destroy/cdk destroy are blocked unless you asked for the specific stack/config key=value 문법 추가 (예: /config thinking=false) — 대화형·-p·Remote Control 모두 동작
sandbox.allowAppleEvents opt-in 설정 추가
CLAUDE_CLIENT_PRESENCE_FILE 환경변수 추가 — 마커 파일을 지정하면 PC 앞에 있을 때 모바일 푸시 알림 억제
Improved N memories line to no longer list individual files outside verbose modeopen, osascript, and browser-based auth flows failing with error -600 on macOS by adding the Apple Events entitlementawsCredentialExport credentials with a short remaining lifetime causing credential refreshes every minute, and now accepts the JSON shape from aws configure export-credentialsdenyRead/allowRead glob over a large directory tree making the Bash tool description enormous and the session unusable on LinuxTool(param:value) 문법 추가 — 툴 입력 파라미터 매칭(* 와일드카드), 예: Agent(model:opus)로 Opus 서브에이전트 차단
.claude/skills 디렉터리의 스킬도 그 안 파일 작업 시 로드 — 이름 충돌 시 중첩 스킬은 <dir>:<name>로 표시돼 둘 다 유지
.claude/ 디렉터리: 이름 충돌 시 작업 폴더에 가장 가까운 agent·workflow·output-style이 우선 — 프로젝트 범위 워크플로 저장도 가장 가까운 .claude/workflows/로
/bug는 이제 제출 전 설명 필수 — 모델 거부 문구를 GitHub 이슈 제목으로 쓰지 않음
/doctor with consistent flat tree layout across all sections, clearer section status icons, and highlighted command names.claude/skills skills with directory-qualified names being blocked by permission prompts in non-interactive runsclaude agents workers failing with 401 Invalid bearer token when the daemon was started from a shell with a custom API gateway via ANTHROPIC_BASE_URL and ANTHROPIC_AUTH_TOKENlanguage 설정으로 특정 언어 고정)
footerLinksRegexes 설정 추가 (사용자·관리 설정)
claude daemon status가 버전 불일치 동작 설명
awsCredentialExport are now cached until their Expiration instead of a fixed 1 houravailableModels enforcement: alias model picks can no longer be redirected to a blocked model via ANTHROPIC_DEFAULT_*_MODEL environment variables, and /fast now refuses to toggle when it would switch to a model outside the allowlist.claude/settings.json is a symlink with an absolute targetenforceAvailableModels 관리 설정 추가 — 켜면 availableModels 허용목록이 Default 모델까지 제한(허용 안 된 모델로 풀리는 Default는 첫 허용 모델로 폴백), 사용자·프로젝트 설정으로 관리 목록을 넓힐 수 없음
wheelScrollAccelerationEnabled 설정 추가
/usage)에 사용량 출처 표시 — 캐시 미스·롱컨텍스트·서브에이전트, 스킬/에이전트/플러그인/MCP별 분해(최근 24h·7d)
/advisor dialog pre-selecting a saved advisor model that is blocked by the availableModels allowlistAWS_REGION 미설정 시 ~/.aws 설정에서 리전 읽음(AWS SDK 우선순위 일치), /status가 리전 출처 표시
/plugin에서 마켓플레이스 플러그인 탐색 시 검색바 추가
claude_code.lines_of_code.count OTEL 메트릭에 model 속성 추가
/code-review가 claude.ai 미로그인 시에도 ultra 옵션 노출(클라우드 리뷰엔 claude.ai 계정 필요 안내)
/loop 권유 중단(대기 루프가 컨테이너를 살려두지 못함)
/goal status chip no longer re-renders the terminal at 5 Hz while idle, and fewer UI re-renders while subagents run in parallelavailableModels restrictions not being applied to subagent model overrides, the agent dispatch model picker, and the advisor modelavailableModels allowlists hiding the /model picker's Opus and Sonnet 1M rows when entries use version-specific IDs like claude-opus-4-8WebFetch(domain:*.example.com) wildcard domain rules never matching subdomains in allow, deny, and ask position, and file permission rules with mid-pattern wildcards (e.g. Read(secrets-*/config.json)) being rejected at startup--safe-mode 플래그(및 CLAUDE_CODE_SAFE_MODE) 추가 — CLAUDE.md·플러그인·스킬·훅·MCP 전부 끄고 시작(문제 해결용)
/cd 명령 추가 — 세션 중 프롬프트 캐시 안 깨고 작업 디렉터리 이동
disableBundledSkills 설정·CLAUDE_CODE_DISABLE_BUNDLED_SKILLS 환경변수 추가 — 번들 스킬·워크플로·내장 슬래시 명령을 모델에서 숨김
/workflows 즉시 열림
API_FORCE_IDLE_TIMEOUT=0으로 해제
--ide·--chrome·--bare·--remote-control 등 플래그 보존, respawn 상태 검증 강화
EnterWorktree 전 헛된 거부 편집 방지
claude.exe를 다른 프로세스가 잡고 있으면 세션 내 재시도 중단
claude agents 권유 팁 추가
TaskCreate reliability: malformed inputs are repaired automatically and validation errors for unloaded tools include the schemaallowedMcpServers/deniedMcpServers) not being enforced on reconnect, IDE-typed configs, --mcp-config servers during the first session after install, or before remote settings loaded; also fixed slow cold starts for orgs without remote settingsfallbackModel 설정 추가 — 기본 모델 과부하·불가 시 최대 3개를 순서대로 폴백, --fallback-model도 대화형 세션에 적용
"*"는 전체 차단), allow 규칙은 비-MCP glob 거부, deny의 미지 툴명은 시작 시 경고
SendMessage로 중계한 메시지는 사용자 권한 없음(수신측이 중계된 권한 요청 거부, auto 모드는 차단)
MAX_THINKING_TOKENS=0·--thinking disabled·모델별 thinking 토글이 Claude API 기본 thinking 모델의 thinking을 끔(3P 제공자는 동일)
claude update가 다운로드 전 대상 버전을 알림(조용히 진행하지 않음)
claude agents: 목록에 URL 입력 시 첫 프롬프트에 그 URL이 있던 세션으로 필터
/login to clear a stale auth check after toggling /voicerequiredMinimumVersion·requiredMaximumVersion 관리 설정 추가 — 버전이 허용 범위 밖이면 시작 거부하고 승인 버전 안내
/plugin list 명령 추가(--enabled/--disabled 필터)
/btw에 "c로 복사" 단축키 추가 — 원시 마크다운 답을 클립보드에 복사(붙여넣기 시 서식 유지)
hookSpecificOutput.additionalContext 반환 가능 — 훅 오류로 처리되지 않고 Claude에 피드백 주며 턴 이어감
$를 넣는 \$ 이스케이프 문법 추가
--resume 시 훅/Bash와 동일한 CLAUDE_CODE_SESSION_ID 받음
claude agents 상태별 뷰에서 디스패치 시 에이전트 뷰를 연 디렉터리에서 세션 시작
$TMPDIR was overridden to /tmp/claude-{uid} for all commands instead of only sandboxed ones (regression in 2.1.154)claude agents --json에 waitingFor 추가 — 대기 세션이 무엇에 막혔는지 표시(예: 권한 프롬프트)
--tools: Grep/Glob 명시 시 임베디드 검색 네이티브 빌드에서 전용 검색 툴 제공(이전엔 무시됨)
/effort가 선택 레벨을 새 세션 기본값으로 유지할지 확인
/ide 메뉴·/terminal-setup·/scroll-speed에서 Windsurf를 Devin Desktop으로 개명(에디터 리브랜드 반영)
claude update verification to wait out endpoint-security scanning of new binaries instead of failing after 5 secondsWebFetch(domain:...) deny/ask/allow rules now take precedence over the preapproved-host auto-allow~\, \\server\share) or case-variant paths, and Read deny rules not hiding files from Glob/Grep resultsOTEL_RESOURCE_ATTRIBUTES 값을 메트릭 데이터포인트 라벨로 포함 → 팀·repo 같은 커스텀 차원으로 사용량 분석
claude agents 행이 fan-out 시 상세 앞에 done/total 표시, peek는 가장 오래 실행 항목 표시
/mcp가 한 번도 로그인 안 한 claude.ai 커넥터를 "미사용 커넥터 표시" 행 뒤로 접음
wl-copy/xclip/xsel 사용(가능 시), 클립보드와 PRIMARY 선택 둘 다 복사(중클릭 붙여넣기), "네이티브 선택하려면 {key}" 힌트가 터미널별 올바른 키 표시
/terminal-setup) 권유 팁 추가
forceLoginOrgUUID/forceLoginMethod managed-settings policies blocking third-party provider sessions (Bedrock, Vertex, Foundry, Mantle) alongside the org pin (regression in 2.1.146)claude mcp list/get/add printing secrets to the terminal: ${VAR} references are no longer expanded, and credential headers and URL secrets are redactedisolation: "worktree" in background sessions being blocked from editing files inside their own worktree.zshenv·.zlogin·.bash_login)과 ~/.config/git/ 쓰기 전 확인 프롬프트 추가(의도치 않은 명령 실행 방지)
acceptEdits 모드가 코드 실행 권한을 주는 빌드도구 설정파일(.npmrc·.yarnrc*·bunfig.toml·.bazelrc·.pre-commit-config.yaml·.devcontainer/ 등) 쓰기 전 확인
grep으로 파일 본 뒤 Edit에 별도 Read 불필요 — 단일 파일 grep/egrep/fgrep이 read-before-edit 충족
CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE 제거(이제 no-op)
workflow에서 ultracode로 변경. "workflow"는 더 이상 트리거 안 함(자기 말로 요청하면 동작), 트리거 키워드는 입력창에서 보라색 강조
claude agentsclaude rm/stop, idle reap) to send SIGTERM to running shell subprocesses before SIGKILL, so cleanup handlers runCLAUDE_CODE_ENABLE_AUTO_MODE=1으로 활성화
.claude/skills 디렉터리의 플러그인이 마켓플레이스 없이 자동 로드
.claude/skills에 새 플러그인 스캐폴딩하는 claude plugin init <name> 추가
/plugin 인수 자동완성 추가 — 서브커맨드·설치된 플러그인 이름·알려진 마켓플레이스 플러그인
claude agents: 디스패치된 세션에 settings.json의 agent 필드 적용, --agent <name>으로 재정의 가능
EnterWorktree가 세션 중 Claude 관리 worktree 간 전환 가능
OTEL_LOG_TOOL_DETAILS=1 시 tool_decision 텔레메트리 이벤트에 tool_parameters(bash 명령·MCP/스킬 이름) 포함
git worktree remove/prune으로 정리 가능
alt+v 단축키)·Windows 11 스크린샷 붙여넣기 수정, Windows 탐색기에서 이미지 드래그 지원 추가
/terminal-setup이 VS Code/Cursor/Windsurf 통합 터미널에서 GPU 가속 비활성화 — 텍스트 깨짐 방지
claude agents: 디스패치 입력의 슬래시 명령 자동완성이 부분 문자열 매칭
/status와 명령 차단 시 계속 표시
claude agents completed sessions not retiring when an idle subagent was still parked or had leaked a backgrounded shell/workflows로 실행 현황 확인
/simplify가 전체 버그 탐지 리뷰 대신 정리 전용 리뷰(재사용·단순화·효율·고도) 실행 후 수정 적용
/effort 슬라이더 레이블을 "Speed"/"Intelligence"에서 "Faster"/"Smarter"로 변경
claude agents: ! <command> 입력으로 셸 명령을 attach/detach 가능한 백그라운드 세션으로 실행. claude --bg --exec '<command>'로도 사용 가능
claude agents: /logout이 백그라운드 세션으로 전달되지 않고 로그아웃 처리
←←가 Bedrock·Vertex·Foundry 및 텔레메트리 비활성화 상태에서도 동작
/chrome → "Select browser…"로 연결된 브라우저 선택, 또는 여러 브라우저 연결 시 채팅 중 선택
plugin.json이나 마켓플레이스 항목에 defaultEnabled: false 선언 가능 — /plugin 또는 claude plugin enable로 활성화. 활성화된 플러그인의 의존성은 자동 활성화
/plugin Discover 탭이 현재 디렉터리와 연관성 높은 플러그인을 "suggested for this directory" 주석과 함께 고정
CLAUDE_CODE_SESSION_ID·CLAUDECODE=1 수신
claude mcp list/get가 파이프 출력 시 자동 승인·연결 대신 미승인 .mcp.json 서버를 ⏸ Pending approval로 표시
/remote-control 자동완성에 "Disconnect Remote Control" 표시
/claude-api 스킬에 Claude Opus 4.8 지원 및 4.7 → 4.8 마이그레이션 안내 추가
CLAUDE_CODE_OPUS_4_6_FAST_MODE_OVERRIDE deprecated(06/01 제거 예정). Opus 4.6 빠른 모드는 /model claude-opus-4-6[1m] 전환 후 /fast on 사용
claude.exe 사용 중 업데이트 실패 시 일반 오류 대신 다른 세션 닫고 재시도 안내
$TMPDIR resolving to different directories in sandboxed vs unsandboxed Bash commands within the same sessionallowedMcpServers/deniedMcpServers entry in managed settings discarding all managed-settings policy; the bad entry is now dropped with a claude doctor warninggithub/git 플러그인 마켓플레이스 소스에 clone·업데이트 시 Git LFS 다운로드를 건너뛰는 skipLfs 옵션 추가
/doctor에서 수정 방법 제공
COLUMNS·LINES 환경변수 수신 — 스크립트가 터미널 너비에 맞게 출력 크기 조정 가능
claude agents: 디스패치 입력 자동완성이 프로젝트 스킬뿐 아니라 기본 슬래시 명령·번들 스킬도 제안
claude agents: PR 컬럼이 단일 PR은 PR #N, 복수 PR은 N PRs로 표시
claude doctor now shows the result of your last update attempt
--strict-mcp-config no longer strips inline mcpServers from explicitly-passed agent definitions (--agents / SDK agents), and blocked subagent MCP servers now surface a visible warning
/bg while Claude is responding now continues the response in the background session instead of dropping it
/model now saves your selection as the default for new sessions (matching the IDE). Press s in the picker to switch models for the current session only.
modelPicker:setAsDefault keybinding, rename it to modelPicker:thisSessionOnly in keybindings.json (the d action was replaced by s)
$CLAUDE_JOB_DIR triggering a "sensitive file" permission prompt/code-review --fix now applies review findings to your working tree after the review, surfacing reuse, simplification, and efficiency suggestions; /simplify now invokes /code-review --fix
disallowed-tools in frontmatter to remove tools from the model while the skill is active
/reload-skills command to re-scan skill directories without restarting the session
SessionStart hooks can now return reloadSkills: true to re-scan skill directories, making skills installed by the hook available in the same session
SessionStart hooks can now set the session title via hookSpecificOutput.sessionTitle on startup and resume
MessageDisplay hook event that lets hooks transform or hide assistant message text as it is displayed
pluginSuggestionMarketplaces managed setting: admins can allowlist org marketplaces whose plugins may be suggested via context-aware tips
claude plugin marketplace remove now accepts --scope user|project|local for symmetry with marketplace add, install, and uninstall
--fallback-model for the rest of the session when the primary model is not found, instead of failing every request
/ in NORMAL mode now opens reverse history search (like Ctrl+R), matching bash/zsh vi-mode
/usage breakdown now includes large session files; files are scanned with a streaming read so memory usage stays flat
Ctrl+O shows the full thinking)
app.entrypoint, opt-in via OTEL_METRICS_INCLUDE_ENTRYPOINT=true)
claude agents when accepting a stale permission prompt after a subagent was cancelled/usage 카테고리 분석
/usage
skills·subagents·plugins·MCP별 비용. 어디서 토큰이 새는지 파악.
/diff 키보드 스크롤
화살표·j/k·Space·PgUp/PgDn
diff 뷰에서 마우스 없이 키보드로 완전 탐색.
- [x] / - [ ] 시각적 표시.
cd·pushd·popd 후 오래된 변수 추적값을 파서가 신뢰하던 갭 수정.
cd 함수가 디렉토리 변경 탐지 못해 workspace 외부 파일 접근 가능하던 취약점 수정.
.git 대신 전체 메인 저장소 루트를 커버하던 문제 수정.
/code-review (구 /simplify)
/code-review [--effort high] [--comment]
effort 수준 선택 후 정확성 버그 리뷰. --comment로 GitHub PR 인라인 코멘트 게시.
forceLoginOrgUUID·forceLoginMethod가 서드파티·API 키 세션에 시행 안 되던 문제 수정.claude agents --json
$ claude agents --json
스크립팅용 JSON 세션 목록. tmux-resurrect·상태 표시줄·세션 선택기 활용.
agent_id/parent_agent_id
subagent spans이 디스패칭 Agent tool span 아래에 중첩. 분산 추적 구조 개선.
/plugin Discover 미리보기/plugin → Discover설치 전 commands·agents·skills·hooks·MCP/LSP 서버 목록 확인 가능./resume 백그라운드 세션
/resume → bg 마크 세션 선택
claude --bg·agent view 세션이 bg 마크와 함께 목록에 표시.
/model 세션별 설정
/model → d키로 기본값 고정
현재 세션에만 모델 변경. d 키로 새 세션 기본값 설정 가능.
worktree.bgIsolation: "none"
settings.json에 추가
백그라운드 세션이 worktree 없이 직접 작업 복사본 편집 가능.
-ExecutionPolicy Bypass 기본값CLAUDE_CODE_POWERSHELL_RESPECT_EXECUTION_POLICY=1PowerShell 도구가 기본으로 -ExecutionPolicy Bypass 전달. 환경변수로 옵트아웃.claude agents flags
--model --effort --permission-mode --mcp-config 등
--add-dir·--settings·--mcp-config·--plugin-dir·--permission-mode·--model·--effort·--dangerously-skip-permissions
/goal 목표 기반 자율 실행
/goal "완료 조건 입력"
완료 조건 설정 후 Claude가 여러 턴 자율 진행. 오버레이에 경과·턴·토큰 실시간 표시.
terminalSequence hook 필드
stop hook에 terminalSequence 추가
터미널 없이 데스크톱 알림·창 제목·벨 출력. 백그라운드 작업 완료 감지에 활용.
claude agents --cwd
$ claude agents --cwd /path
세션 목록을 특정 디렉토리로 범위 지정.
CLAUDE_CODE_PLUGIN_PREFER_HTTPS
GitHub SSH 키 없는 환경에서 HTTPS로 플러그인 클론.
ANTHROPIC_WORKSPACE_ID
워크로드 ID 페더레이션용 환경 변수.
subagent_type 매칭대소문자·구분자 무관 매칭. "Code Reviewer" → code-reviewer 자동 처리./scroll-speed
/scroll-speed → 슬라이더
마우스 휠 스크롤 속도를 실시간 미리보기로 조정.
? 단축키 도움말, {/}로 프롬프트 간 이동, v로 패널 토글.
args 필드
hook 설정에 args: ["cmd","arg1"] 추가
exec form으로 셸 없이 직접 명령 실행. 경로 플레이스홀더 따옴표 불필요.
CLAUDE_PROJECT_DIR
MCP stdio 서버가 환경에서 CLAUDE_PROJECT_DIR 수신.
/schedule·MCP 커넥터 제한ANTHROPIC_API_KEY/apiKeyHelper/ANTHROPIC_AUTH_TOKEN 설정 시 Remote Control·/schedule·claude.ai MCP 커넥터 기본 비활성화.CLAUDE_CODE_ENABLE_FEEDBACK_SURVEY_FOR_OTEL to re-enable the session quality survey for enterprises capturing responses through OpenTelemetry
settings.autoMode.hard_deny for auto mode classifier rules that block unconditionally regardless of user intent or allow exceptions
d (matching delete elsewhere) instead of r which collided with retry
--worktree collides with an existing or stale worktreeworktree.baseRef setting (fresh | head) to choose whether --worktree, EnterWorktree, and agent-isolation worktrees branch from origin/<default> or local HEAD. Note: the default fresh changes EnterWorktree's base back to origin/<default> (it has been local HEAD since 2.1.128) — set worktree.baseRef: "head" to keep unpushed commits in new worktrees
sandbox.bwrapPath and sandbox.socatPath managed settings (Linux/WSL) to specify custom bubblewrap and socat binary locations
parentSettingsBehavior admin-tier key ('first-wins' | 'merge') to let admins opt SDK managedSettings (parent tier) into the policy merge
effort.level JSON input field and the $CLAUDE_EFFORT environment variable, and Bash tool commands can read $CLAUDE_EFFORT
claude --help now lists --remote-control alongside --remote-control-session-name-prefix
claudeCode.claudeProcessWrapper failing with "Unsupported platform" when the extension build doesn't bundle a Claude binary
HTTP(S)_PROXY / NO_PROXY / mTLS not being respected for the full MCP OAuth flow including discovery, dynamic client registration, token exchange, and token refreshCLAUDE_CODE_SESSION_ID environment variable to the Bash tool subprocess environment, matching the session_id passed to hooks
CLAUDE_CODE_DISABLE_ALTERNATE_SCREEN=1 env var to opt out of the fullscreen alternate-screen renderer and keep the conversation in the terminal's native scrollback
/tui fullscreen startup banner to describe additional renderer benefits (lower memory usage, mouse support, auto-copy on select)
/login, /upgrade, /extra-usage dialog spacing--permission-mode flag being ignored when resuming a plan-mode session with -p --continue/--resume, and plan mode not being re-applied after ExitPlanMode within the same session-p mode retrying non-transient 4xx connection failuresx-api-key header--plugin-url <url> flag to fetch a plugin .zip archive from a URL for the current session
CLAUDE_CODE_FORCE_SYNC_OUTPUT=1 env var to force-enable synchronized output on terminals that auto-detection misses (e.g. Emacs eat)
CLAUDE_CODE_PACKAGE_MANAGER_AUTO_UPDATE: when set on Homebrew or WinGet installations, Claude Code runs the upgrade command in the background and prompts to restart
themes and monitors should now be declared under "experimental": { ... }. Top-level declarations still work but claude plugin validate will warn
/v1/models discovery for the /model picker is now opt-in via CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY=1 (was automatic in 2.1.126–2.1.128)
ANTHROPIC_BASE_URL gateway) no longer see spinner tips pointing at first-party Anthropic surfaces
skillOverrides setting now works: off hides from model and /, user-invocable-only hides from model only, name-only collapses description
claude_code.pull_request.count OTel metric now counts PRs/MRs created via MCP tools, not just shell commands
/clear not clearing the conversation context and displayed transcript
/rename disappearing while a permission or other dialog is activeuser:inference scope/color (no args) now picks a random session color
/mcp now shows the tool count for connected servers and flags servers that connected with 0 tools
--plugin-dir now accepts .zip plugin archives in addition to directories
--channels now works with console (API key) authentication — console orgs with managed settings must set channelsEnabled: true to enable
/model picker: collapsed duplicate Opus 4.7 entries, and current Opus now shows as "Opus" instead of "Opus 4.7"
OTEL_* environment variables, so OTEL-instrumented apps run via the Bash tool no longer pick up the CLI's own OTLP endpoint
workspace is now a reserved server name — existing servers with that name will be skipped with a warning
localSettings suggestion for Bash permission prompts, so "Always allow" writes to .claude/settings.local.json
EnterWorktree now creates the new branch from local HEAD as documented, instead of origin/<default-branch> — unpushed commits are no longer dropped
/compact, or run with --debug)
--output-format stream-json: init.plugin_errors now includes --plugin-dir load failures in addition to dependency demotions
/model picker now lists models from your gateway's /v1/models endpoint when ANTHROPIC_BASE_URL points at an Anthropic-compatible gateway
claude project purge [path] to delete all Claude Code state for a project (transcripts, tasks, file history, config entry) — supports --dry-run, -y/--yes, -i/--interactive, and --all
--dangerously-skip-permissions now bypasses prompts for writes to .claude/, .git/, .vscode/, shell config files, and other previously-protected paths (catastrophic removal commands still prompt as a safety net)
claude auth login now accepts the OAuth code pasted into the terminal when the browser callback can't reach localhost (WSL2, SSH, containers)
claude_code.skill_activated OpenTelemetry event now fires for user-typed slash commands and carries a new invocation_trigger attribute ("user-slash", "claude-proactive", or "nested-skill")
CLAUDE_CODE_PROVIDER_MANAGED_BY_HOST) no longer auto-disable analytics on Bedrock/Vertex/Foundry
.NET global tool is now detected
allowManagedDomainsOnly / allowManagedReadPathsOnly being ignored when a higher-priority managed-settings source lacked a sandbox block
-- (e.g. git diff -- file) is no longer mis-flagged as the --% stop-parsing token
CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS=1 is setANTHROPIC_BEDROCK_SERVICE_TIER environment variable to select a Bedrock service tier (default, flex, or priority), sent as the X-Amzn-Bedrock-Service-Tier header
/resume search box now finds the session that created that PR (GitHub, GitHub Enterprise, GitLab, and Bitbucket)
/mcp now shows claude.ai connectors hidden by a manually-added server with the same URL, with a hint to remove the duplicate
/mcp message shown when an MCP server is still unauthorized after the browser sign-in flow
api_request/api_error log events are now emitted as numbers, not strings
claude_code.at_mention log event for @-mention resolution
alwaysLoad option to MCP server config — when true, all tools from that server skip tool-search deferral and are always available
claude plugin prune to remove orphaned auto-installed plugin dependencies; plugin uninstall --prune cascades
/skills so you can find a skill in long lists without scrolling
hookSpecificOutput.updatedToolOutput (previously MCP-only)
claude -p: CLAUDE_CODE_FORK_SUBAGENT=1 now works in non-interactive sessions
--dangerously-skip-permissions no longer prompts for writes to .claude/skills/, .claude/agents/, and .claude/commands/
/terminal-setup now enables iTerm2's "Applications in terminal may access clipboard" setting so /copy works, including from tmux
language setting
mcp_authenticate now supports redirectUri for custom scheme completion and claude.ai connectors
stop_reason, gen_ai.response.finish_reasons, and user_system_prompt (gated behind OTEL_LOG_USER_PROMPTS) to LLM request spans
accessibility.voice.speechLanguage setting when no Claude Code language is configured
/context now opens a native token usage dialog
find in the Bash tool on large directory trees/usage leaking up to ~2GB of memory on machines with large transcript historiesprompt parameter/usage returning "rate limited" after a stale OAuth token — now refreshes automatically/config settings (theme, editor mode, verbose, etc.) now persist to ~/.claude/settings.json and participate in project/local/policy override precedence
prUrlTemplate setting to point the footer PR badge at a custom code-review URL instead of github.com
CLAUDE_CODE_HIDE_CWD environment variable to hide the working directory in the startup logo
--from-pr now accepts GitLab merge-request, Bitbucket pull-request, and GitHub Enterprise PR URLs
--print mode now honors the agent's tools: and disallowedTools: frontmatter, matching interactive-mode behavior
--agent <name> now honors the agent definition's permissionMode for built-in agents
PostToolUse and PostToolUseFailure hook inputs now include duration_ms (tool execution time, excluding permission prompts and PreToolUse hooks)
owner/repo#N shorthand links in output now use your git remote's host instead of always pointing at github.com
blockedMarketplaces now correctly enforces hostPattern and pathPattern entries
tool_result and tool_decision events now include tool_use_id; tool_result also includes tool_input_size_bytes
effort.level and thinking.enabled
ENABLE_TOOL_SEARCH)
--client-secret not being sent during token exchange for servers requiring client_secret_postisolation: "worktree" reusing stale worktrees from prior sessionsv) and visual-line mode (V) with selection, operators, and visual feedback
/cost and /stats into /usage — both remain as typing shortcuts that open the relevant tab
/theme, or hand-edit JSON files in ~/.claude/themes/; plugins can also ship themes via a themes/ directory
type: "mcp_tool"
DISABLE_UPDATES env var to completely block all update paths including manual claude update — stricter than DISABLE_AUTOUPDATER
wslInheritsWindowsSettings policy key
"$defaults" in autoMode.allow, autoMode.soft_deny, or autoMode.environment to add custom rules alongside the built-in list instead of replacing it
claude plugin tag to create release git tags for plugins with version validation
--continue/--resume now find sessions that added the current directory via /add-dir
/color now syncs the session accent color to claude.ai/code when Remote Control is connected
/model picker now honors ANTHROPIC_DEFAULT_*_MODEL_NAME/_DESCRIPTION overrides when using a custom ANTHROPIC_BASE_URL gateway
/doctor and the /plugin Errors tab
/mcp menu hiding OAuth Authenticate/Re-authenticate actions for servers configured with headersHelper, and HTTP/SSE MCP servers with custom headers being stuck in "needs authentication" after a transient 401expires_in requiring re-authentication every hourinsufficient_scope 403 names a scope the current token already has~/.claude/.credentials.json/login having no effect in a session launched with CLAUDE_CODE_OAUTH_TOKEN — the env token is now cleared so disk credentials take effect--dangerously-skip-permissionsCLAUDE_CODE_FORK_SUBAGENT=1
mcpServers are now loaded for main-thread agent sessions via --agent
/resume command now offers to summarize stale, large sessions before re-reading them, matching the existing --resume behavior
plugin install on an already-installed plugin now installs any missing dependencies instead of stopping at "already installed"
claude plugin marketplace add now auto-resolves missing dependencies from configured marketplaces
blockedMarketplaces and strictKnownMarketplaces are now enforced on plugin install, update, refresh, and autoupdate
/compact
cleanupPeriodDays retention sweep now also covers ~/.claude/tasks/, ~/.claude/shell-snapshots/, and ~/.claude/backups/
user_prompt events now include command_name and command_source for slash commands; cost.usage, token.usage, api_request, and api_error now include an effort attribute when the model supports effort levels. Custom/MCP command names are redacted unless OTEL_LOG_TOOL_DETAILS=1 is set
Glob and Grep tools are replaced by embedded bfs and ugrep available through the Bash tool — faster searches without a separate tool round-trip (Windows and npm-installed builds unchanged)
where.exe executable lookups per process for faster subprocess launches
high (was medium)
/model: selections now persist across restarts even when the project pins a different model, and the startup header shows when the active model comes from a project or managed-settings pin/login having no effect when launched with CLAUDE_CODE_OAUTH_TOKEN env var and that token expires/resume on large sessions is significantly faster (up to 67% on 40MB+ sessions) and handles sessions with many dead-fork entries more efficiently
/terminal-setup now configures the editor's scroll sensitivity
/config search now matches option values (e.g. searching "vim" finds the Editor mode setting)
/doctor can now be opened while Claude is responding, without waiting for the current turn to finish
/reload-plugins and background plugin auto-update now auto-install missing plugin dependencies from marketplaces you've already added
gh commands hit GitHub's API rate limit, so agents can back off instead of retrying
hooks: now fire when running as a main-thread agent via --agent
rm/rmdir targeting /, $HOME, or other critical system directories
https://downloads.claude.ai/claude-code-releases instead of https://storage.googleapis.com/claude-code-dist-86c565f3-f756-42ad-8dfa-d59b1c096819/claude-code-releases
resources/templates/list is now deferred to first @-mentionsandbox.network.deniedDomains setting to block specific domains even when a broader allowedDomains wildcard would otherwise permit them
Ctrl+A and Ctrl+E now move to the start/end of the current logical line in multiline input, matching readline behavior
Ctrl+Backspace now deletes the previous word
/extra-usage now works from Remote Control (mobile/web) clients
@-file autocomplete suggestions
cd <current-directory> && git … no longer triggers a permission prompt when the cd is a no-op
/private/{etc,var,tmp,home} paths are now treated as dangerous removal targets under Bash(rm:*) allow rules
env/sudo/watch/ionice/setsid and similar exec wrappers
Bash(find:*) allow rules no longer auto-approve find -exec/-delete
/loop: pressing Esc now cancels pending wakeups, and wakeups display as "Claude resuming /loop wakeup" for clarity/ultrareview: faster launch with parallelized checks, diffstat in the launch dialog, and animated launching statedangerouslyDisableSandbox running commands outside the sandbox without a permission promptxhigh effort level for Opus 4.7, sitting between high and max. Available via /effort, --effort, and the model picker; other models fall back to high
/effort now opens an interactive slider when called without arguments, with arrow-key navigation between levels and Enter to confirm
/theme
/less-permission-prompts skill — scans transcripts for common read-only Bash and MCP tool calls and proposes a prioritized allowlist for .claude/settings.json
/ultrareview for running comprehensive code review in the cloud using parallel multi-agent analysis and critique — invoke with no arguments to review your current branch, or /ultrareview <PR#> to fetch and review a specific GitHub PR
CLAUDE_CODE_USE_POWERSHELL_TOOL. On Linux and macOS, enable with CLAUDE_CODE_USE_POWERSHELL_TOOL=1 (requires pwsh on PATH)
ls *.ts) and commands starting with cd <project-dir> && no longer trigger a permission prompt
claude <word> is invoked with a near-miss typo (e.g. claude udpate → "Did you mean claude update?")
fix-auth-race-snug-otter.md) instead of purely random words
/skills menu now supports sorting by estimated token count — press t to toggle
Ctrl+U now clears the entire input buffer (previously: delete to start of line); press Ctrl+Y to restore
Ctrl+L now forces a full screen redraw in addition to clearing the prompt input
[ (dump to scrollback) and v (open in editor) shortcuts
--output-format stream-json now includes plugin_errors on the init event when plugins are demoted for unsatisfied dependencies
OTEL_LOG_RAW_API_BODIES environment variable to emit full API request and response bodies as OpenTelemetry log events for debugging
CLAUDE_ENV_FILE and SessionStart hook environment files now apply (previously a no-op)
--enable-auto-mode/setup-vertex and /setup-bedrock to show the actual settings.json path when CLAUDE_CONFIG_DIR is set, seed model candidates from existing pins on re-run, and offer a "with 1M context" option for supported modelsplugin update; plugin install now recovers from interrupted prior installs/tui command and tui setting — run /tui fullscreen to switch to flicker-free rendering in the same conversation
Ctrl+O to toggle between normal and verbose transcript only; focus view is now toggled separately with the new /focus command
autoScrollEnabled config to disable conversation auto-scroll in fullscreen mode
Ctrl+G external editor (enable via /config)
--resume/--continue now resurrects unexpired scheduled tasks
/context, /exit, and /reload-plugins now work from Remote Control (mobile/web) clients
TRACEPARENT/TRACESTATE from the environment for distributed trace linking
DISABLE_TELEMETRY). Opt out via /config or CLAUDE_CODE_ENABLE_AWAY_SUMMARY=0.
/plugin Installed tab — items needing attention and favorites appear at the top, disabled items are hidden behind a fold, and f favorites the selected item/doctor to warn when an MCP server is defined in multiple config scopes with different endpointsPermissionRequest hooks returning updatedInput not being re-checked against permissions.deny rules; setMode:'bypassPermissions' updates now respect disableBypassPermissionsModeENABLE_PROMPT_CACHING_1H env var to opt into 1-hour prompt cache TTL on API key, Bedrock, Vertex, and Foundry (ENABLE_PROMPT_CACHING_1H_BEDROCK is deprecated but still honored), and FORCE_PROMPT_CACHING_5M to force 5-minute TTL
/config and manually invocable with /recap; force with CLAUDE_CODE_ENABLE_AWAY_SUMMARY if telemetry disabled.
/init, /review, and /security-review via the Skill tool
/undo is now an alias for /rewind
Ctrl+O)
DISABLE_PROMPT_CACHING* environment variables
/model to warn before switching models mid-conversation, since the next response re-reads the full history uncached/resume picker to default to sessions from the current directory; press Ctrl+A to show all projectspath parameter to the EnterWorktree tool to switch into an existing worktree of the current repository
{"decision":"block"}
monitors manifest key that auto-arms at session start or on skill invoke
/proactive is now an alias for /loop
/doctor layout with status icons; press f to have Claude fix reported issues/config labels and descriptions for clarityWebFetch to strip <style> and <script> contents from fetched pages so CSS-heavy pages no longer exhaust the content budget before reaching actual textjq for JSON, computed Read chunk sizes for text)acceptEdits permission mode when exiting plan mode would downgrade from a higher permission level/team-onboarding command to generate a teammate ramp-up guide from your local Claude Code usage
CLAUDE_CODE_CERT_STORE=bundled to use only bundled CAs)
/ultraplan and other remote-session features now auto-create a default cloud environment instead of requiring web setup first
claude -p --resume <name> to accept session titles set via /rename or --namesettings.json no longer causes the entire file to be ignoredallowManagedHooksOnly is set/plugin and claude plugin update to show a warning when the marketplace could not be refreshed, instead of silently reporting a stale versionOTEL_LOG_USER_PROMPTS, OTEL_LOG_TOOL_DETAILS, and OTEL_LOG_TOOL_CONTENT; sensitive span attributes are no longer emitted unless opted inquery() to clean up subprocess and temp files when consumers break from for await or use await usingpermissions.deny rules not overriding a PreToolUse hook's permissionDecision: "ask" — previously the hook could downgrade a deny into a promptANTHROPIC_AUTH_TOKEN, apiKeyHelper, or ANTHROPIC_CUSTOM_HEADERS set an Authorization headermktemp: No such file or directory after a fresh boot/mcp menu offering OAuth-specific actions for MCP servers configured with headersHelper; Reconnect is now offered instead to re-invoke the helper script/login OAuth URL rendering with padding that prevented clean mouse selectionCLAUDE_CODE_PERFORCE_MODE env var: when set, Edit/Write/NotebookEdit fail on read-only files with a p4 edit hint instead of silently overwriting them
CLAUDE_CODE_SUBPROCESS_ENV_SCRUB is set, and CLAUDE_CODE_SCRIPT_CAPS env var to limit per-session script invocations
--exclude-dynamic-system-prompt-sections flag to print mode for improved cross-user prompt caching
workspace.git_worktree to the status line JSON input, set whenever the current directory is inside a linked git worktree
TRACEPARENT env var to Bash tool subprocesses when OTEL tracing is enabled, so child-process spans correctly parent to Claude Code's trace tree
clientInfo in the initialize request
/claude-api skill to cover Managed Agents alongside Claude API
CLAUDE_CODE_GIT_BASH_PATH is set or Git is installed at a default location
/compact hints when DISABLE_COMPACT is set.
/resume filter hint labels and added project/worktree/branch names in the filter indicator/agents with a tabbed layout: a Running tab shows live subagents, and the Library tab adds Run agent and View running instance actions/reload-plugins to pick up plugin-provided skills without requiring a restartj/k in NORMAL mode now navigate history and select the footer pill at the input boundary--debugoauth.authServerMetadataUrl config override not being honored on token refresh after restart, affecting ADFS and similar IdPs--dangerously-skip-permissions being silently downgraded to accept-edits mode after approving a write to a protected path via Bashpermissions.additionalDirectories changes not applying mid-session — removed directories lose access immediately and added ones work without restartBash(cmd:*) and Bash(git commit *) wildcard permission rules failing to match commands with extra spaces or tabscut -d /, paste -d /, column -s /, awk '{print $1}' file, and filenames containing %toString) causing settings.json to be silently ignored--dangerously-skip-permissions_meta["anthropic/maxResultSizeChars"] not bypassing the token-based persist layersandbox.network.allowMachLookup not taking effect on macOSCtrl+O) in NO_FLICKER mode showing prompt, one-line tool summary with edit diffstats, and final response
refreshInterval status line setting to re-run the status line command every N seconds
workspace.git_worktree to the status line JSON input, set when the current directory is inside a linked git worktree
● N running indicator in /agents next to agent types with live subagent instances
.cedar, .cedarpolicy)
/claude-api skill to cover Managed Agents alongside the Claude API
LANG=C rm foo, timeout 5 mkdir out)sandbox.network.allowMachLookup now takes effect on macOS@-mention completion to trigger after CJK sentence punctuation, so Japanese/Chinese input no longer requires a space before / or @TRACEPARENT env var when tracing is enabled--dangerously-skip-permissions being silently downgraded to accept-edits mode after approving a write to a protected pathtoString) causing settings.json to be silently ignoredpermissions.additionalDirectories changes in settings not applying mid-sessionsettings.permissions.additionalDirectories revoking access to the same directory passed via --add-diroauth.authServerMetadataUrl not being honored on token refresh after restart, fixing ADFS and similar IdPscwd: override leaking their working directory back to the parent session's Bash toolNO_FLICKER mode memory leak where API retries left stale streaming stateAWS_BEARER_TOKEN_BEDROCK or ANTHROPIC_BEDROCK_BASE_URL are set to empty strings (as GitHub Actions does for unset inputs)403 "Authorization header is missing" when using AWS_BEARER_TOKEN_BEDROCK or CLAUDE_CODE_SKIP_BEDROCK_AUTH (regression in 2.1.94)CLAUDE_CODE_USE_MANTLE=1
/effort)
Slacked #channel header with a clickable channel link for Slack MCP send-message tool calls
keep-coding-instructions frontmatter field support for plugin output styles
hookSpecificOutput.sessionTitle to UserPromptSubmit hooks for setting the session title
"skills": ["./"] now use the skill's frontmatter name for the invocation name instead of the directory basename, giving a stable name across install methods
settings.json files fail to parse, so users know their permission rules are not being applied
--resume to resume sessions from other worktrees of the same repo directly instead of printing a cd commandforceRemoteSettingsRefresh policy setting: when set, the CLI blocks startup until remote managed settings are freshly fetched, and exits if the fetch fails (fail-closed)
/cost for subscription users
/release-notes is now an interactive version picker
myhost-graceful-unicorn), overridable with --remote-control-session-name-prefix
/tag command
/vim command (toggle vim mode via /config → Editor mode)
apply-seccomp helper in both npm and native builds, restoring unix-socket blocking for sandboxed commands
&/$)_meta["anthropic/maxResultSizeChars"] annotation (up to 500K), allowing larger results like DB schemas to pass through without truncation
disableSkillShellExecution setting to disable inline shell execution in skills, custom slash commands, and plugin commands
claude-cli://open?q= deep links (encoded newlines %0A no longer rejected)
bin/ and invoke them as bare commands from the Bash tool
/feedback now explains why it's unavailable instead of disappearing from the slash menu
old_string anchors, reducing output tokens
/claude-api skill guidance for agent design patterns including tool surface decisions, context management, and caching strategystripAnsi on Bun by routing through Bun.stripANSIpermissions.defaultMode: "auto" in settings.json/powerup — interactive lessons teaching Claude Code features with animated demos
CLAUDE_CODE_PLUGIN_KEEP_MARKETPLACE_ON_FAILURE env var to keep the existing marketplace cache when git pull fails, useful in offline environments
.husky to protected directories (acceptEdits mode)
& background job bypass, -ErrorAction Break debugger hang, archive-extraction TOCTOU, and parse-fail fallback deny-rule degradation
--resume picker to no longer show sessions created by claude -p or SDK invocations
Get-DnsClientCache and ipconfig /displaydns from auto-allow (DNS cache privacy)
/resume all-projects view to load project sessions in parallel, improving load times for users with many projects"defer" permission decision to PreToolUse hooks — headless sessions can pause at a tool call and resume with -p --resume to have the hook re-evaluate
CLAUDE_CODE_NO_FLICKER=1 environment variable to opt into flicker-free alt-screen rendering with virtualized scrollback
PermissionDenied hook that fires after auto mode classifier denials — return {retry: true} to tell the model it can retry
@ mention typeahead suggestions
MCP_CONNECTION_NONBLOCKING=true for -p mode to skip the MCP connection wait entirely, and bounded --mcp-config server connections at 5s instead of blocking on the slowest server
/permissions → Recent tab where you can retry with r
Edit to work on files viewed via Bash with sed -n or cat, without requiring a separate Read call first
cleanupPeriodDays: 0 in settings.json to be rejected with a validation error — it previously silently disabled transcript persistence
showThinkingSummaries: true in settings.json to restore
TaskCreated hook event and its blocking behavior
/env now applies to PowerShell tool commands (previously only affected Bash)
/usage now hides redundant "Current week (Sonnet only)" bar for Pro and Enterprise plans
!command into an empty prompt now enters bash mode, matching typed ! behavior
/buddy is here for April 1st — hatch a small creature that watches you code
ls/tree/du instead of "Read N files"@-mention typeahead to rank source files above MCP resources with similar namesX-Claude-Code-Session-Id header to API requests so proxies can aggregate requests by session without parsing the body
.jj and .sl to VCS directory exclusion lists so Grep and file autocomplete don't descend into Jujutsu or Sapling metadata
/skills listing are now capped at 250 characters to reduce context usage
/skills menu to sort alphabetically for easier scanning
@ — raw string content no longer JSON-escapedc shortcut copying only ~20 characters of the OAuth login URL instead of the full URLCLAUDE_CODE_MCP_SERVER_NAME and CLAUDE_CODE_MCP_SERVER_URL environment variables to MCP headersHelper scripts, allowing one helper to serve multiple servers
if field for hooks using permission rule syntax (e.g., Bash(git *)) to filter when they run, reducing process spawning overhead
/loop, CronCreate) fire
[Image #N] placeholder when pasting images
claude-cli://open?q=…) now support up to 5,000 characters, with a "scroll to review" warning for long pre-filled prompts
managed-settings.json) can no longer be installed or enabled, and are hidden from marketplace views
AskUserQuestion by returning updatedInput alongside permissionDecision: "allow", enabling headless integrations that collect answers via their own UI
tool_parameters in OpenTelemetry tool_result events are now gated behind OTEL_LOG_TOOL_DETAILS=1
403 insufficient_scope now correctly trigger the re-authorization flowANTHROPIC_DEFAULT_{OPUS,SONNET,HAIKU}_MODEL_SUPPORTS env vars to override effort/thinking capability detection for pinned default models for 3p (Bedrock, Vertex, Foundry), and _MODEL_NAME/_DESCRIPTION to customize the /model picker label
CLAUDE_STREAM_IDLE_TIMEOUT_MS env var to configure the streaming idle watchdog threshold (default 90s)
TaskCreated hook that fires when a task is created via TaskCreate
WorktreeCreate hook support for type: "http" — return the created worktree path via hookSpecificOutput.worktreePath in the response JSON
allowedChannelPlugins managed setting for team/enterprise admins to define a channel plugin allowlist
x-client-request-id header to API requests for debugging timeouts
/clear, reducing unnecessary token re-caching on stale sessions
claude-cli://) now open in your preferred terminal instead of whichever terminal happens to be first in the detection list
paths: frontmatter now accepts a YAML list of globs
ToolSearch is enabled, including for users with MCP tools configured
owner/repo#123 — bare #123 is no longer auto-linked
/voice, /mobile, /chrome, /upgrade, etc.) are now hidden instead of shown
C:\, C:\Windows, etc.)setup() in parallel with slash command and agent loadingclaude "prompt" with MCP servers — the REPL now renders immediately instead of blocking until all servers connectEdit(.claude) allow rulesmanaged-settings.d/ drop-in directory alongside managed-settings.json, letting separate teams deploy independent policy fragments that merge alphabetically
CwdChanged and FileChanged hook events for reactive environment management (e.g., direnv)
sandbox.failIfUnavailable setting to exit with an error when sandbox is enabled but cannot start, instead of running unsandboxed
disableDeepLinkRegistration setting to prevent claude-cli:// protocol handler registration
CLAUDE_CODE_SUBPROCESS_ENV_SCRUB=1 to strip Anthropic and cloud provider credentials from subprocess environments (Bash tool, hooks, MCP stdio servers)
/ in transcript mode (Ctrl+O) to search, n/N to step through matches
Ctrl+X Ctrl+E as an alias for opening the external editor (readline-native binding; Ctrl+G still works)
[Image #N] chip at the cursor so you can reference them positionally in your prompt
initialPrompt in frontmatter to auto-submit a first turn
chat:killAgents and chat:fastMode are now rebindable via ~/.claude/keybindings.json
/status now works while Claude is responding, instead of being queued until the turn finishes
XDG_DATA_HOME when registering the claude-cli:// protocol handler
Ctrl+F to Ctrl+X Ctrl+K to stop shadowing readline forward-char
TaskOutput tool in favor of using Read on the background task's output file path
CLAUDE_CODE_DISABLE_NONSTREAMING_FALLBACK env var to disable the non-streaming fallback when streaming fails
manifest.userConfig) now available externally — plugins can prompt for configuration at enable time, with sensitive: true values stored in keychain (macOS) or protected credentials file (other platforms)
Ctrl+L now clears the screen and forces a full redraw — use this to recover when Cmd+K leaves the UI partially blank. Use Ctrl+U or double-Esc to clear prompt input.
MEMORY.md index now truncates at 25KB as well as 200 lines
AskUserQuestion and plan-mode tools when --channels is active
/rewind) to open a keyboard-navigable rewind picker
--resume memory usage and startup latency on large sessionsWebFetch to identify as Claude-User so site operators can recognize and allowlist Claude Code traffic via robots.txtWebFetch peak memory usage for large pagesclaude -p startup with unauthenticated HTTP/SSE MCP servers (~600ms saved)--bare -p (SDK pattern) is ~14% faster to the API request--mcp-config CLI flag bypassing allowedMcpServers/deniedMcpServers managed policy enforcementwhich is kernel-restrictedrg ... | wc -l and similar piped commands hanging and returning 0 in sandbox mode on Linux--bare flag for scripted -p calls — skips hooks, LSP, plugin sync, and skill directory walks; requires ANTHROPIC_API_KEY or an apiKeyHelper via --settings (OAuth and keychain auth disabled); auto-memory fully disabled
--channels permission relay — channel servers that declare the permission capability can forward tool approval prompts to your phone
"showClearContextOnPlanAccept": true)
! bash mode discoverability — Claude now suggests it when you need to run an interactive command--channels bypass for Team/Enterprise orgs with no other managed settings configuredrate_limits field to statusline scripts for displaying Claude.ai rate limit usage (5-hour and 7-day windows with used_percentage and resets_at)
source: 'settings' plugin marketplace source — declare plugin entries inline in settings.json
effort frontmatter support for skills and slash commands to override the model effort level when invoked
--channels (research preview) — allow MCP servers to push messages into your session
@ file autocomplete in large git repositories/effort to show what auto currently resolves to, matching the status bar indicator/permissions — Tab and arrow keys now switch tabs from within a list/plugin install command instead of a two-step flow/sandbox tab switching not responding to Tab or arrow keysenabledPlugins, permissions.defaultMode, policy-set env vars) not being applied at startup when remote-settings.json was cached from a prior session--console flag to claude auth login for Anthropic Console (API billing) authentication
/config menu
CLAUDE_CODE_PLUGIN_SEED_DIR now supports multiple seed directories separated by the platform path delimiter (: on Unix, ; on Windows)
/remote-control — bridge your session to claude.ai/code to continue from a browser or phone
/permissionsStopFailure hook event that fires when the turn ends due to an API error (rate limit, auth failure, etc.)
${CLAUDE_PLUGIN_DATA} variable for plugin persistent state that survives plugin updates; /plugin uninstall prompts before deleting it
effort, maxTurns, and disallowedTools frontmatter support for plugin-shipped agents
set -g allow-passthrough on
sandbox.enabled: true is set but dependencies are missing — now shows a visible startup warning
ANTHROPIC_CUSTOM_MODEL_OPTION env var to add a custom entry to the /model picker, with optional _NAME and _DESCRIPTION suffixed vars for display
git log HEAD failing with "ambiguous argument" inside sandboxed Bash on Linux, and stub files polluting git status in the working directorydeny: ["mcp__servername"] permission rules not removing MCP server tools before sending to the model, allowing it to see and attempt blocked toolssandbox.filesystem.allowWrite not working with absolute paths (previously required // prefix)/sandbox Dependencies tab showing Linux prerequisites on macOS instead of macOS-specific info.git, .claude, and other protected directories being writable without a prompt in bypassPermissions modeallowRead sandbox filesystem setting to re-allow read access within denyRead regions
/copy now accepts an optional index: /copy N copies the Nth-latest assistant response
apiKeyHelper takes longer than 10s, preventing it from blocking the main loop
resume parameter — use SendMessage({to: agentId}) to continue a previously spawned agent
SendMessage now auto-resumes stopped agents in the background instead of returning an error
/fork to /branch (/fork still works as an alias)
macOptionClickForcesSelection setting
--resume on fork-heavy and very large sessions — up to 45% faster loading and ~100-150MB less peak memoryclaude plugin validate to check skill, agent, and command frontmatter plus hooks/hooks.json, catching YAML parse errors and schema violationsCLAUDE_CODE_PLUGIN_SEED_DIRcd src && npm test) saving a single rule for the full string instead of per-subcommand, leading to dead rules and repeated permission prompts"allow" bypassing deny permission rules, including enterprise managed settings←/→ accidentally switching tabs in settings, permissions, and sandbox dialogs while navigating listsElicitation and ElicitationResult hooks to intercept and override responses before they're sent back
-n / --name <name> CLI flag to set a display name for the session at startup
worktree.sparsePaths setting for claude --worktree in large monorepos to check out only the directories you need via git sparse-checkout
PostCompact hook that fires after compaction completes
/effort slash command to set model effort level
feedbackSurveyRate setting
--plugin-dir to only accept one path to support subcommands — use repeated --plugin-dir for multiple directories
--worktree startup performance by reading git refs directly and skipping redundant git fetch when the remote branch is already available locally/voice to show your dictation language on enable and warn when your language setting isn't supported for voice inputBash(cmd:*) permission rules not matching when a quoted argument contains #/color command for all users to set a prompt-bar color for your session
/rename
--verbose or transcript mode)
C:\ProgramData\ClaudeCode\managed-settings.json — use C:\Program Files\ClaudeCode\managed-settings.json
/context command — identifies context-heavy tools, memory bloat, and capacity warnings with specific optimization tips
autoMemoryDirectory setting to configure a custom directory for auto-memory storage
--plugin-dir so local dev copies now override installed marketplace plugins with the same name (unless that plugin is force-enabled by managed settings)
ask rules being bypassed by user allow rules or skill allowed-toolsaudio-input entitlement so macOS prompts correctlymodelOverrides setting to map model picker entries to custom provider model IDs (e.g. Bedrock inference profile ARNs)
NODE_EXTRA_CA_CERTS)
/output-style command — use /config instead. Output style is now fixed at session start for better prompt caching
/effort to work while Claude is responding, matching /model behaviorw key in /copy to write the focused selection directly to a file, bypassing the clipboard (useful over SSH)
/plan (e.g., /plan fix the auth bug) that enters plan mode and immediately starts
ExitWorktree tool to leave an EnterWorktree session
CLAUDE_CODE_DISABLE_CRON environment variable to immediately stop scheduled cron jobs mid-session
lsof, pgrep, tput, ss, fd, and fdfind to the bash auto-approval allowlist, reducing permission prompts for common read-only operations
model parameter on the Agent tool for per-invocation model overrides
<!-- ... -->) to be hidden from Claude when auto-injected. Comments remain visible when read with the Read tool
.git suffix (Azure DevOps, AWS CodeCommit)
vscode://anthropic.claude-code/open URI handler to open a new Claude Code tab programmatically, with optional prompt and session query parameters
/effort auto to reset to default/config — Escape now cancels changes, Enter saves and closes, Space toggles settingsworktreePath and worktreeBranch/tmp/claude/) prompting unnecessarilysandbox.excludedCommands failing with env var prefixes; "always allow" suggesting overly broad prefixes for nested CLI tools; and deny rules not applying to all command forms/loop command to run a prompt or slash command on a recurring interval (e.g. /loop 5m check the deploy)
voice:pushToTalk keybinding to make the voice activation key rebindable in keybindings.json (default: space) — modifier+letter combos like meta+k have zero typing interference
fmt, comm, cmp, numfmt, expr, test, printf, getconf, seq, tsort, and pr to the bash auto-approval allowlist
/debug to toggle debug logging on mid-session, since debug logs are no longer written by default
/plugin uninstall to disable project-scoped plugins in .claude/settings.local.json instead of modifying .claude/settings.json, so changes don't affect teammates/plugin menu./permissions Workspace tab when the same directory is added with and without a trailing slash/mcp in the chat panel to enable/disable servers, reconnect, and manage OAuth authentication without switching to the terminal
/rename to work while Claude is processing, instead of being silently queued/poll rate to once per 10 minutes while connected (was 1–2s), cutting server load ~300×. Reconnection is unaffected — transport loss immediately wakes fast polling.permissions.defaultMode settings values other than acceptEdits or plan being applied in Claude Code Remote environments — they are now ignored/claude-api skill for building applications with the Claude API and Anthropic SDK
!) to exit bash mode, matching escape and backspace
/remote-control and claude remote-control (/remote-control My Project or --name "My Project") to set a custom session title visible in claude.ai/code
claude --agent
sandbox.enableWeakerNetworkIsolation setting (macOS only) to allow Go programs like gh, gcloud, and terraform to verify TLS certificates when using a custom MITM proxy with httpProxyPort
includeGitInstructions setting (and CLAUDE_CODE_DISABLE_GIT_INSTRUCTIONS env var) to remove built-in commit and PR workflow instructions from Claude's system prompt
/reload-plugins command to activate pending plugin changes without restarting
${CLAUDE_SKILL_DIR} variable for skills to reference their own directory in SKILL.md content
InstructionsLoaded hook event that fires when CLAUDE.md or .claude/rules/*.md files are loaded into context
agent_id (for subagents) and agent_type (for subagents and --agent) to hook events
worktree field to status line hook commands with name, path, branch, and original repo directory when running in a --worktree session
pluginTrustMessage in managed settings to append organization-specific context to the plugin trust warning shown before installation
pathPattern to strictKnownMarketplaces for regex-matching file/directory marketplace sources alongside hostPattern restrictions
git-subdir to point to a subdirectory within a git repo
oauth.authServerMetadataUrl config option for MCP servers to specify a custom OAuth metadata discovery URL when standard discovery fails
/resume picker to show your most recent prompt instead of the first one. This also resolves some titles appearing as (session).
activeForm field — the spinner falls back to the task subject
permissions.disableBypassPermissionsMode from your effective Claude Code settings (including managed/policy settings) — when set to disable, bypass permissions mode is hidden from the picker
--worktree startup by eliminating a git subprocess on the startup path-p startup by pipelining claude.ai config fetch with local connections and using a concurrency pool instead of sequential batchingonSubmit across message updates--append-system-prompt-file and --system-prompt-file work in interactive mode (the docs previously said print mode only)security -i stdin buffer, silently leaving stale credentials behind and causing repeated /login prompts..credentials.json losing subscriptionType (showing "Claude API" instead of "Claude Pro"/"Claude Max") when the profile endpoint transiently fails during token refresh (anthropics/claude-code#30185).bashrc, HEAD, etc.) appearing as untracked files in the working directory after sandboxed Bash commands on LinuxacceptEdits modeallowManagedDomainsOnly is enabled in managed settings — non-allowed domains are now blocked automatically with no bypassAskUserQuestion) being silently auto-allowed when listed in a skill's allowed-tools, bypassing the permission prompt and running with empty answersmemoCache over long sessionscd <outside-dir> && <cmd> permission prompt to surface the chained command instead of only showing "Yes, allow reading from <dir>/"[ERROR] logs with 403s from profile-scoped endpoints/model
/simplify and /batch bundled slash commands
ENABLE_CLAUDEAI_MCP_SERVERS=false env var to opt out from making claude.ai MCP servers available
/copy picker. When selected, future /copy commands will skip the code block picker and copy the full response directly.
/model command to show the currently active model in the slash command menu/copy command to show an interactive picker when code blocks are present, allowing selection of individual code blocks or the full response.
cd /tmp && git fetch && git push) to compute smarter per-subcommand prefixes instead of treating the whole command as oneclaude remote-control subcommand for external builds, enabling local environment serving for all users.
CLAUDE_CODE_PLUGIN_GIT_TIMEOUT_MS to configure.
-l flag) by default when a shell snapshot is available, improving command execution performance. Previously this required setting CLAUDE_BASH_NO_LOGIN=true.
CLAUDE_CODE_ACCOUNT_UUID, CLAUDE_CODE_USER_EMAIL, and CLAUDE_CODE_ORGANIZATION_UUID environment variables for SDK callers to provide account info synchronously, eliminating a race condition where early telemetry events lacked account metadata.
/model picker now shows human-readable labels (e.g., "Sonnet 4.5") instead of raw model IDs for pinned model versions, with an upgrade hint when a newer version is available.
startupTimeout configuration for LSP servers
WorktreeCreate and WorktreeRemove hook events, enabling custom VCS setup and teardown when agent worktree isolation creates or removes worktrees.
isolation: worktree in agent definitions, allowing agents to declaratively run in isolated git worktrees.
CLAUDE_CODE_SIMPLE mode now also disables MCP tools, attachments, hooks, and CLAUDE.md file loading for a fully minimal experience.
claude agents CLI command to list all configured agents
CLAUDE_CODE_DISABLE_1M_CONTEXT environment variable to disable 1M context window support
/extra-usage command support in VS Code sessions
-p flag) by deferring Yoga WASM and UI component imports--worktree (-w) flag to start Claude in an isolated git worktree
isolation: "worktree" for working in a temporary git worktree
background: true to always run as a background task
settings.json for default configuration
CLAUDE_CODE_SIMPLE) now includes the file edit tool in addition to the Bash tool, allowing direct file editing in simple mode.
supportsEffort, supportedEffortLevels, and supportsAdaptiveThinking fields so consumers can discover model capabilities.
ConfigChange hook event that fires when configuration files change during a session, enabling enterprise security auditing and optional blocking of settings changes.
-p) by skipping unnecessary API calls during startupdisableAllHooks setting to respect managed settings hierarchy — non-managed settings can no longer disable managed hooks set by policy (#26637)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.
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.
chat:newline keybinding action for configurable multi-line input (anthropics/claude-code#26075)
added_dirs to the statusline JSON workspace section, exposing directories added via /add-dir to external scripts (anthropics/claude-code#26096)
~/.claude/backups/ to reduce home directory clutter (anthropics/claude-code#26130)
/rename command now updates the terminal tab title by default (anthropics/claude-code#25789)
@ file mentions - file suggestions now appear faster by pre-warming the index on startup and using session-based caching with background refresh.enabledPlugins and extraKnownMarketplaces from --add-dir directories
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
SDKRateLimitInfo and SDKRateLimitEvent types to the SDK, enabling consumers to receive rate limit status updates including utilization, reset times, and overage information
speed attribute to OTel events and trace spans for fast mode visibility
claude auth login, claude auth status, and claude auth logout CLI subcommands
/rename to auto-generate session name from conversation context when called without arguments.claude/skills directory in sandbox mode
sandbox.excludedCommands or dangerouslyDisableSandbox) could bypass the Bash ask permission rule when autoAllowBashIfSandboxed was enabledTeammateIdle and TaskCompleted hook events for multi-agent workflows
Task(agent_type) syntax in agent "tools" frontmatter
memory frontmatter field support for agents, enabling persistent memory with user, project, or local scope
/skills menu for better discoverability
.claude/skills/ within additional directories (--add-dir) are now loaded automatically.
cat, sed, grep, find), reducing unnecessary bash command usagepages parameter to the Read tool for PDFs, allowing specific page ranges to be read (e.g., pages: "1-5"). Large PDFs (>10 pages) now return a lightweight reference when @ mentioned instead of being inlined into context.
--client-id and --client-secret with claude mcp add.
/debug for Claude to help troubleshoot the current session
git log and git show flags in read-only mode (e.g., --topo-order, --cherry-pick, --format, --raw)
/model to execute immediately instead of being queued
--resume (68% reduction for users with many sessions) by replacing the session index with lightweight stat-based loading and progressive enrichmentTaskStop tool to display the stopped command/task description in the result line instead of a generic "Task stopped" message--from-pr flag to resume sessions linked to a specific GitHub PR number or URL
gh pr create
.bashrc files
spinnerVerbs)
python and pip commands use the correct interpreter (configurable via claudeCode.usePythonEnvironment setting)CLAUDE.md files from additional directories specified via --add-dir flag (requires setting CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD=1)TaskUpdate toolToolSearch results to appear as a brief notification instead of inline in the conversation/commit-push-pr skill to automatically post PR URLs to Slack channels when configured via MCP tools/copy command to be available to all usersBash(*) to be accepted and treated as equivalent to Bash/sandbox command UI to show dependency status with installation instructions when dependencies are missingCLAUDE_CODE_ENABLE_TASKS, set to false to keep the old system temporarily$0, $1, etc. for accessing individual arguments in custom commands$ARGUMENTS.0 to $ARGUMENTS[0] (bracket syntax)queued_command attachment messages as SDKUserMessageReplay events when replayUserMessages is enabled/keybindings to get started. Learn more at https://code.claude.com/docs/en/keybindingsclaude install or see https://docs.anthropic.com/en/docs/claude-code/getting-started for more options!) - type a partial command and press Tab to complete from your bash command history/usage command to display current plan usageSetup hook event that can be triggered via --init, --init-only, or --maintenance CLI flags for repository setup and maintenance operationsauto:N syntax for configuring the MCP tool search auto-enable threshold, where N is the context window percentage (0-100)plansDirectory setting to customize where plan files are storedPreToolUse hooks to return additionalContext to the model${CLAUDE_SESSION_ID} string substitution for skills to access the current session IDshowTurnDuration setting to hide turn duration messages (e.g., "Cooked for 1m 6s")MCPSearch to disallowedTools in their settings.claudeProcessWrapper setting passing the wrapper path instead of the Claude binary path/config command for quickly filtering settings/doctor showing auto-update channel and available npm versions (stable/latest)/stats command - press r to cycle between Last 7 days, Last 30 days, and All time.claude/skills directories when working with files in subdirectoriescontext_window.used_percentage and context_window.remaining_percentage fields to status line input for easier context window display/mcp enable <name> instead/tasks dialog to go directly to task details when there's only one background task runningCLAUDE_CODE_TMPDIR environment variable to override the temp directory used for internal temp files, useful for environments with custom temp directory requirementsCLAUDE_CODE_DISABLE_BACKGROUND_TASKS environment variable to disable all background task functionality including auto-backgrounding and the Ctrl+B shortcutstable or latest) toggle to /config/doctor and after saving rules that include the source of each rule and actionable fix guidanceFORCE_AUTOUPDATE_PLUGINS environment variable to allow plugin autoupdate even when the main auto-updater is disabledagent_type to SessionStart hook input, populated if --agent is specified/plugins installed tab to unify plugins and MCPs with scope-based groupingC:\ProgramData\ClaudeCode\managed-settings.json - administrators should migrate to C:\Program Files\ClaudeCode\managed-settings.jsonscp instead of the unhelpful clipboard shortcut hint~/.claude/skills or .claude/skills are now immediately available without restarting the sessioncontext: fork in skill frontmatteragent field in skills to specify agent type for executionlanguage setting to configure Claude's response language (e.g., language: "japanese")respectGitignore support in settings.json for per-project control over @-mention file picker behaviorIS_DEMO environment variable to hide email and organization from the UI, useful for streaming or recording sessions* at any position in rules (e.g., Bash(npm *), Bash(* install), Bash(git * main))list_changed notifications, allowing MCP servers to dynamically update their available tools, prompts, and resources without requiring reconnection/teleport and /remote-env slash commands for claude.ai subscribers, allowing them to resume and configure remote sessionsTask(AgentName) syntax in settings.json permissions or the --disallowedTools CLI flag; and , to repeat f/F/t/T motions, y operator for yank with yy/Y, p/P for paste, text objects (iw, aw, iW, aW, i", a", i', a', i(, a(, i[, a[, i{, a{), >> and << for indent/dedent, and J to join lines/plan command shortcut to enable plan mode directly from the prompt/ appears anywhere in input, not just at the beginning--tools flag support in interactive mode to restrict which built-in tools Claude can use during interactive sessionsCLAUDE_CODE_FILE_READ_MAX_OUTPUT_TOKENS environment variable to override the default file read token limitonce: true config for hooksallowed-tools field for cleaner skill declarationscat refactor.md | claude/skills/ directories to be visible in the slash command menu by default (opt-out with user-invocable: false in frontmatter)/stats outputupdatedInput when returning ask permission decision, enabling hooks to act as middleware while still requesting user consent/terminal-setup support for Kitty, Alacritty, Zed, and Warp terminals/theme to toggle syntax highlighting on/off/context command visualization with grouped skills and agents by source, slash commands, and sorted token count[Image #N] links that open attached images in the default viewer--session-id combined with --resume or --continue and --fork-session/theme command to open theme picker directly.ignore or .rgignore files/settings as an alias for the /config commandANTHROPIC_BEDROCK_BASE_URL is now respected for token counting and inference profile listing.mcp.json not loading when using --dangerously-skip-permissionsls *.txt, for f in *.png)mcp__server__* for MCP tool permissions to allow or deny all tools from a servercurrent_usage field to status line input, enabling accurate context window percentage calculations/permissions command with / keyboard shortcut for filtering rules by tool name/doctor/permissionsclaude install --force not bypassing stale lock filesfileSuggestion setting for custom @ file search commandsCLAUDE_CODE_SHELL environment variable to override automatic shell detection (useful when login shell differs from actual working shell)/rename to name sessions, /resume <name> in REPL or claude --resume <name> from the terminal to resume themaws login AWS Management Console credentials/resume screen with grouped forked sessions and keyboard shortcuts for preview (P) and rename (R)attribution setting to customize commit and PR bylines (deprecates includeCoAuthoredBy)agent setting to configure main thread with a specific agent's system prompt, tool restrictions, and modelC:\Program Files\ClaudeCode if it exists. Support for C:\ProgramData\ClaudeCode will be removed in a future version.@ file suggestions with faster, more accurate results$! in bash commands/usage back to "% used"claude --teleport/usagePermissionRequest hook to automatically approve or deny tool permission requests with custom logic&permissionMode field for custom agentstool_use_id field to PreToolUseHookInput and PostToolUseHookInput typesSubagentStart hook eventagent_id and agent_transcript_path fields to SubagentStop hooks.model parameter to prompt-based stop hooks, allowing users to specify a custom model for hook evaluationkeep-coding-instructions option to frontmatterCLAUDE_CODE_EXIT_AFTER_STOP_DELAY environment variable to automatically exit SDK mode after a specified idle duration, useful for automated workflows and scriptsignorePatterns from project config to deny permissions in the localSettings.chat.fontSize and chat.fontFamily settings throughout the entire UI, and apply font changes immediately without requiring reloadcompanyAnnouncements setting for displaying announcements on startupsecurity unlock-keychain when encountering API key errors on macOS with locked keychainallowUnsandboxedCommands sandbox setting to disable the dangerouslyDisableSandbox escape hatch at policy leveldisallowedTools field to custom agent definitions for explicit tool blocking/output-style and use --system-prompt-file, --system-prompt, --append-system-prompt, CLAUDE.md, or plugins instead/terminal-setup from adding backslash to Shift + Enter in VS Codeowner/repo#branch)structuredContent field in tool responsesANTHROPIC_DEFAULT_HAIKU_MODELCLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC now disables release notes fetching/plugin install, /plugin enable/disable, /plugin marketplace commands for plugin managementextraKnownMarketplaces for team collaboration/plugin validate command for validating plugin structure and configuration/doctor command/model selector/help/resume summaries/config now persist across sessionsglobal.anthropic.claude-sonnet-4-5-20250929-v1:0--agents flagv0.69.0 → v0.145.0 · 2025.12.10 – 2026.7.21
stable 릴리즈만 (alpha 제외) · 출처: GitHub releases
멀티에이전트 V2 정식화 — 서브에이전트 모델·reasoning·동시성·역할 설정 가능, 에이전트 네비게이션 개선.
Cursor·Claude Code 설정·MCP·플러그인·세션·명령·메모리를 한 번에 가져오기.
Code mode에서 웹 검색 직접 호출 가능 — JS 중첩 툴 호출 포함, 결과는 일반 텍스트로.
세션 아카이브로 실수 재개·포크 방지. codex unarchive로 복원.
명명된 권한 프로파일 인식·설정된 커스텀 프로파일 표시.
Vim 모드에 텍스트 오브젝트 편집·word/line-end 동작 개선·인터럽트 턴 바인딩 설정 추가.
패키지 Codex 빌드가 번들된 패치 zsh 헬퍼를 macOS·Linux에서 발견·사용.
Python SDK가 스레드·턴 API용 친숙한 Sandbox 프리셋 노출.
TUI 입력 프롬프트가 비활성 시 자동 확정 — 카운트다운은 상호작용 시 일시정지.
플러그인이 스레드별 stdio MCP 서버 활성화 가능 — "내가 만든" 마켓플레이스·인증별 카탈로그 추가.
/usage에서 획득한 사용 한도 리셋 크레딧 조회·차감 가능 — 확인 후 적용.
롤아웃 토큰 예산으로 에이전트 스레드 전체 사용량 추적·잔여 알림·소진 시 자동 중단.
인덱싱 웹 검색 모드 — 실시간 검색 허용, 직접 페이지 접근은 서버 승인 URL만.
지원 시 MCP 도구가 tool search를 기본 사용 — 도구 발견성 향상, 구형 모델·프로바이더 호환성 유지.
원격 플러그인 기본 활성화 — 풍부한 카탈로그·npm 마켓플레이스·원격/로컬 버전 표시.
MCP 도구 tool search 기본 적용 + ChatGPT 호스팅 MCP 서버에서 세션 인증 지원.
새 writes 승인 모드 — 선언된 읽기 전용 액션 자동 허용, 쓰기만 프롬프트.
MCP 도구가 실험적 옵트인 없이 대화형 인증 요청 가능.
now reports richer environment, Git, terminal, app-server, and thread inventory diagnostics for support cases.
shows remote connection details and server version when the TUI is connected over a remote transport
now understands named permission profiles and displays configured custom profiles
now runs like a foreground command, waits for readiness, reports machine status, and keeps explicit daemon-style start/stop commands
now accepts --output-schema, so resumed automations can keep session context while still enforcing structured JSON output
mentions now search files, directories, plugins, and skills in one picker, backed by app-server plugin metadata.
now reports reasoning-token usage for programmatic consumers
can now jump directly to a session by ID or name from the TUI
Added persisted /goal workflows with app-server APIs, model tools, runtime continuation, and TUI controls for create, pause, resume, and clear.
Added codex update, configurable TUI keymaps, plan-mode nudges, action-required terminal titles, and active-turn /statusline and /title edits.
Expanded permission profiles with built-in defaults, sandbox CLI profile selection, cwd controls, and active-profile metadata for clients.
Improved plugin workflows with marketplace installation, remote bundle caching, remote uninstall, plugin-bundled hooks, hook enablement state, and external-agent config import.
now supports the prompt-plus-stdin workflow, so you can pipe input and still pass a separate prompt on the command line
now exposes codex.cwd and codex.homeDir, and saved codex.tool(...) / codex.emitImage(...) references keep working across cells.
can now dynamically import local .js and .mjs files, making it easier to reuse workspace scripts from the REPL
Windows sandbox runs can now enforce proxy-only networking with OS-level egress rules, instead of relying on environment variables alone
App-server clients can now start ChatGPT sign-in with a device code flow, which helps when browser callback login is unreliable or unavailable
Custom model providers can now fetch and refresh short-lived bearer tokens dynamically, instead of being limited to static credentials from config or environment variables.
now accepts inline prompt arguments and pasted images, and slash-command editing/highlighting in the TUI is more polished
Added a direct install script for macOS and Linux and publish it as a GitHub release asset, using the existing platform payload (including codex and rg)
Expanded the app-server v2 thread API with experimental thread-scoped realtime endpoints/notifications and a thread/unsubscribe flow to unload live threads without archiving them
Promoted js_repl to /experimental, added startup compatibility checks with user-visible warnings, and lowered the validated minimum Node version to 22.22.0
Enabled request_user_input in Default collaboration mode (not just Plan mode)
Made 5.3-codex visible in the CLI model list for API users
metadata now includes upgrade migration markdown so clients can display richer guidance when suggesting model upgrades
Added an optional SOCKS5 proxy listener with policy enforcement and config gating
Plan mode now streams proposed plans into a dedicated TUI view, plus a feature-gated /plan shortcut for quick mode switching.
Added /apps to browse connectors in TUI and $ insertion for app prompts
App-server can now run in external auth mode, accepting ChatGPT auth tokens from a host app and requesting refreshes when needed
Smart approvals are now enabled by default, with explicit approval prompts for MCP tool calls.
TUI2: Normalize mouse wheel + trackpad scrolling across terminals; add tui.scroll_* config settings (PR #8357)
Add allowed_sandbox_modes to requirements.toml to constrain permitted sandbox modes (PR #8298)
MCP: OAuth login for streamable HTTP MCP servers no longer requires the rmcp_client feature flag (PR #8087)
Improve fuzzy file search display/consistency by centralizing file-name derivation in codex-file-search (PR #8334)
Update bundled model metadata (models.json) (PR #8168)
Fix /undo interacting destructively with git staging / ghost commits (PR #8303)
/import가 Cursor·Claude Code 설정, MCP 서버, 플러그인, 세션, 명령어, 프로젝트 메모리 마이그레이션으로 확장.rm detection, consistent full-access confirmation, and preserved rejection reasons across tools.rm forms, and provides clearer rejection reasons when commands are denied.rust-v0.144.2 이후 병합된 풀 리퀘스트 변경 없음.codex executable.writes 앱 승인 모드 추가.codex_apps connector.codex remote-control pair 추가max 추론 노력 일급 지원respect_system_proxy 활성화 시 macOS 인증 클라이언트가 시스템 프록시·PAC·WPAD 설정 준수/usage에서 적립된 사용량 한도 리셋 크레딧 확인·사용 가능 — 확인·재시도·갱신 상태 포함./plugins가 원격 플러그인을 OpenAI Curated·Workspace·Shared with me 섹션으로 분류, 적합한 턴에서 관련 플러그인 추천·설치 가능.Ctrl+Z and resuming with fg.AGENTS.md discovery, and sandbox behavior across operating systems.thread/list and thread/search.codex exec thread start and resume, while blocking PostToolUse hooks correctly reject code-mode tool calls.wait_agent./usage에 일·주·누적 계정 토큰 사용량 뷰 추가./goal이 대용량 텍스트·붙여넣기 블록·이미지 첨부 보존 — 원격 app-server 세션 포함.codex delete·/delete·app-server thread/delete로 세션 영구 삭제 추가 — 확인 안전장치·서브에이전트 정리 포함./import 추가 — Claude Code에서 설정·프로젝트 구성·최근 대화를 선택 가져오기.@ 입력 시 파일·플러그인·스킬 통합 멘션 메뉴가 기본으로 열림./review from crashing when Esc is pressed with queued guidance, while preserving that guidance when the review is canceled.oneOf·allOf 보존, 큰 스키마는 압축 시 구조 유지 — MCP 툴 호환성↑.codex doctor 로컬 리포트에 에디터·페이저 환경 정보 추가(JSON 출력은 원시값 가림).codex plugin marketplace list --json에 소스 포함, 목록은 캐시 카탈로그 먼저 반환 후 백그라운드 갱신.codex resume --last "..." and codex fork --last "..." now treat the trailing argument as the initial prompt instead of misreading it as a session ID.~ in the path are now linkified end to end in the TUI instead of being truncated before the tilde./new, /clear, and /fork no longer drop cloud-managed requirements or feature flags during TUI config reloads./app 명령으로 현재 CLI 스레드를 macOS·Windows의 Codex Desktop으로 넘김 — Windows는 Desktop으로 바로 열림.Alt 미지원 터미널용 폴백 단축키 추가, 모델이 알린 순서대로 effort 적용.--json 지원, 목록 JSON에 마켓 소스, 상세에 기본 프롬프트·원격 MCP 서버 노출./goal edit no longer submits early, idle auto-turns stay out of Plan mode, and goals stop auto-continuing after terminal turn failures./usr/bin/bash, shorter Linux proxy socket paths, and pre-refresh of expired OAuth-backed MCP credentials.AGENTS.md files are picked up consistently.codex plugin list --json 출력·캐시 원격 카탈로그 제안 추가 (#25330, #25457).skills fields as warnings (#25491, #25681, #25717, #25782)./archive·CLI codex archive/codex unarchive로 세션 아카이브 — 복원 전까지 재개/포크 차단.codex app-server --stdio로 stdio 모드 실행.CODEX_API_KEY 등록 지원, 원격제어 웹소켓은 단명 서버 토큰 사용.codex sandbox setup --elevated 프로비저닝 경로 + 허용 Windows 샌드박스 requirements 지원./diff from running repository-provided Git helpers/hooks, avoids PowerShell parser execution on non-Windows hosts, and rejects browser-origin exec-server websocket handshakes.deny read rules stay enforced for safe-command and approval-bypass paths.AWS_REGION / AWS_DEFAULT_REGION, and unsupported Bedrock GPT service tiers are no longer advertised or sent.codex doctor가 환경·Git·터미널·앱서버·스레드 인벤토리 진단을 풍부하게 리포트(지원 케이스용)./status가 원격 전송 연결 시 원격 연결 정보·서버 버전 표시./permissions가 명명된 권한 프로파일 인식·커스텀 프로파일 표시.Sandbox 프리셋 노출.Ctrl-C handling from unsupported keyboard enhancement setup$ suggestions--profile을 CLI·TUI 권한·샌드박스 전반 기본 프로파일 선택자로, 레거시 config는 마이그레이션 안내로 거부.$ref/$defs 구조 보존, 과대 스키마는 노출 전 압축.readOnlyHint 표시 시 동시 실행 허용.codex remote-control이 포그라운드 명령처럼 동작·준비 대기·머신 상태 보고, 데몬식 start/stop 유지.requirements.toml·런타임 새로고침·Windows 샌드박스 통합 강화.TurnResult 반환.codex exec resume가 --output-schema 수용 → 재개 자동화서 세션 컨텍스트 유지하며 구조화 JSON 출력 강제.name (thread-id) in resume hints, and pasted text works in the picker search box/tmp/...-prefixed pathscodex doctor now detects npm-managed installs correctly, and MSVC release binaries no longer depend on separately installed VC++ runtime DLLs@ 멘션이 파일·디렉터리·플러그인·스킬을 한 피커서 검색(앱서버 플러그인 메타데이터 기반).codex remote-control·런타임 enable/disable API·상태 조회·레지스트리 기반 원격 환경 지원.openai-codex/openai_codex로 이전 — 고정 런타임 생성 타입·동시 턴 라우팅·승인 모드·통합 커버리지.codex doctor 추가./review MCP startup status, /side Esc handling, and network approval history texttaskkill cleanup, and non-queued plugin readscodex remote-control 추가.aws login 프로파일의 AWS 콘솔 로그인 자격증명 사용 가능.view_image가 다중 환경 세션서 선택된 환경 통해 파일 해석.ThreadStore, including threads without local rollout pathsresponse.processed for v2 streams and avoids sending service_tier on API-key compact requestscodex exec startup banner/vim·기본 모드 config·Vim 전용 키맵 컨텍스트./ide 컨텍스트 주입·워크스페이스 인식 /diff./keymap debug로 터미널 키 이벤트 점검./hooks서 훅 탐색·토글, compaction 전후 실행·PreToolUse 컨텍스트 추가 가능; Codex Apps auth·MCP elicitation이 TUI/Guardian 흐름에 노출./copy works better in tmux, Alt+Enter and modified Delete/Backspace keys behave correctly, and Windows typing/paste latency was reducedanimations = false for live rowsbwrap, slow mount probes, symlink-protected paths, and shared /tmp setupssafe.directory, and unsafe Git options more reliably/goal 워크플로 추가.codex update·설정 가능 TUI 키맵·plan-mode 넛지·action-required 터미널 타이틀·활성 턴 /statusline·/title 편집 추가.git -C approval handlingapply_patch, GPT-5.4 reasoning levels, and updated Bedrock GPT-5.4 endpoint/model metadatacodex exec --json이 프로그램 소비자용 reasoning-token 사용량 보고./review and exiting the TUI no longer leaves the interface wedged on delegate startup or unsubscribeStart-Process calls avoid visible PowerShell windowsjs_repl image MIME typesAlt+, 낮춤·Alt+. 높임, 모델 업그레이드 수용 시 reasoning을 새 모델 기본값으로 리셋.config.toml·관리형 requirements.toml에 인라인 설정 가능, MCP 툴·apply_patch·장기 Bash 세션 관찰./permissions changes survive side conversations and updated Full Access state is correctly reflected in MCP approval handlingwait_agent so it returns promptly when mailbox work is already queued instead of waiting for a fresh notification or timing outcwd, bringing fallback path resolution in line with CLI behavioramazon-bedrock 모델 제공자 추가./mcp verbose 추가 — 전체 MCP 서버 진단·리소스·리소스 템플릿(/mcp는 빠르게 유지)..mcp.json의 mcpServers·최상위 서버 맵 둘 다 수용.remote_sandbox_config requirements 추가.gpt-5.4 기본 포함./copy after rollback so it copies the latest visible assistant response, not a pre-rollback responseWorking statescodex exec inherit root-level shared flags such as sandbox and model optionscodex app이 Windows·Intel Mac에서 Desktop을 올바르게 열거나 설치/side 대화 가능 — 작업 중 대기 입력도 슬래시 명령·! 셸 프롬프트 지원codex exec 실행 지원js_repl 이미지 출력에 원본 디테일 메타데이터 지원apply_patch writes work correctly with split filesystem policies, and file watchers now notice files created after watching begins/model menu loops, redundant memory notices, and terminal title quoting in iTerm2codex marketplace add + app-server로 GitHub·git URL·로컬 디렉터리·marketplace.json URL에서 플러그인 마켓플레이스 설치 지원Ctrl+R 역방향 검색·수락된 슬래시 명령 로컬 기억danger-full-access denylist-only network moderesume --last and thread/list work when paths use verbatim prefixesprolite plans and made unknown WHAM plan values decodablefs/readDirectoryoutputSchema 상세 포함 → 구조화된 툴 결과 타입 정확도↑./clear로 만들어진 세션을 새 시작·재개 세션과 구별 가능.apply_patch workflowscodex --remote wss://... panics by installing the Rustls crypto provider before TLS websocket connections--cd forwarding, runtime remote-control enablement, sandbox-aware filesystem APIs, and an experimental codex exec-server subcommandCtrl+O, including better clipboard behavior over SSH and across platforms/resume can now jump directly to a session by ID or name from the TUI/status now refreshes stale limits instead of showing frozen or misleading quota information/copy output, percent-decoded local file links, and clearer truncated exec-output hints/fast off in app-server-backed TUI sessions/mcp avoids slow full inventory probes, disabled servers skip auth probing, and residency headers are honored by codex mcp-serverapply_patch errors, refreshed network proxy policy after sandbox changes, suppressed irrelevant bubblewrap warnings, a macOS HTTP-client sandbox panic fix, and Windows firewall address handlingcodex exec now supports the prompt-plus-stdin workflow, so you can pipe input and still pass a separate prompt on the command line.codex files are now protected even on first creation, closing a gap where the initial write could bypass normal approval checksbwrap on normal multi-entry PATHs/copy and /resume <name> work again, /agent no longer shows stale threads, and the skills picker scrolls past the first pageapply_patch is less likely to fail because it no longer adds redundant writable roots that could trigger unnecessary ACL churn/plugins, and install or remove them with clearer auth/setup handling/root/agent_a, with structured inter-agent messaging and agent listing for multi-agent v2 workflows/title terminal-title picker now works in both the classic TUI and the app-server TUI, making parallel sessions easier to tell apart! shell commands, watch filesystem changes, and connect to remote websocket servers with bearer-token authview_image now returns image URLs for code mode, generated images are reopenable from the TUI, and image-generation history survives resumetui_app_server no longer duplicates live reasoning summaries or /review output, and it preserves transcript text instead of dropping it under backpressuretui_app_server now opens the local browser again, cancels cleanly on Ctrl+C, and no longer fails startup when you're logged outShift+Leftbubblewrap, and Windows restricted-token sandboxing now supports more split-policy carveout layoutsuserpromptsubmit hook so prompts can be blocked or augmented before execution and before they enter historyturn/start; startup now times out and falls back cleanlybwrap when availableview_image and codex.emitImage(..., detail: "original"), which helps with precision visual tasksjs_repl now exposes codex.cwd and codex.homeDir, and saved codex.tool(...) / codex.emitImage(...) references keep working across cellscodex tool, with a unified [realtime] session configjs_repl no longer hangs when dynamic tool responses contain literal U+2028 or U+2029 characterscodex exec --profile once again preserves profile-scoped settings when starting or resuming a threadtool_params in approval promptsSessionStart and Stop hook eventsGET /readyz and GET /healthz on the same listener for easier health checks$ mention picker by clearly labeling Skills, Apps, and Plugins, and by surfacing plugins firsttmux crash caused by concurrent user-shell lookupsworkspace-write behavior is preserved and newer permission profiles degrade more safely on older buildsapply_patchrequest_permissions tool so running turns can request additional permissions at runtime, with new TUI rendering for those approval callsplugin/list metadata, install-time auth checks, and a plugin/uninstall endpointexec to the new in-process app server pathgit commands before project trust is establishedTerminateProcess success handling and stricter sandbox startup cwd validation*) domains while preserving scoped wildcard support@plugin mentions so users can reference plugins directly in chat and auto-include their associated MCP/app/skill contextSIGTERM like Ctrl-C for graceful app-server websocket shutdown instead of abrupt termination behaviordata: URLs, preventing external URL forwarding through emitImagejs_repl can now dynamically import local .js and .mjs files, making it easier to reuse workspace scripts from the REPLcodex resume/agent-based enablement, clearer prompts, ordinal nicknames, and role-labeled handoff context/fast toggle in the TUI and app-server support for fast and flex service tiers@ file mentions so parent-directory .gitignore rules no longer hide valid repository files/status, Esc, pending-message handling, and startup/profile race conditionscodex debug clear-memories command to fully reset saved memory state when neededthread/resume now restores pending approval and input requests instead of leaving clients out of syncthread/start no longer blocks unrelated app-server requests, reducing stalls during slow startup paths such as MCP auth checks0.106.0oauth_resource values correctly for servers that require a resource parametercodex and rg)thread/unsubscribe flow to unload live threads without archiving themjs_repl to /experimental, added startup compatibility checks with user-visible warnings, and lowered the validated minimum Node version to 22.22.0request_user_input in Default collaboration mode (not just Plan mode)5.3-codex visible in the CLI model list for API usersCtrl-C handling for sub-agents in the TUI/theme picker with live preview, and uses better theme-aware diff colors for light and dark terminalsspawn_agents_on_csv can fan out work from a CSV with built-in progress/ETA, and sub-agents are easier to follow with nicknames, a cleaner picker, and visible child-thread approval prompts/copy copies the latest complete assistant reply, while /clear and Ctrl-L clear the screen without losing thread context, with /clear also able to start a fresh chatthread/list can search by title, thread status is exposed in read/list responses and notifications, and thread/resume returns the latest turn inline so reconnects are less lossy@ parsing in the chat composer is more reliable, so commands like npx -y @scope/pkg@latest no longer accidentally open the file picker or block submissionpermessage-deflate can connect successfully/dev, fixing failures in tools that need entropy or other standard device nodesjs_repl now reports uncaught kernel failures more clearly, recovers cleanly afterward, and correctly attaches view_image results from nested tool callsWS_PROXY/WSS_PROXY environment support (including lowercase variants) for websocket proxying in the network proxyCtrl+C/Ctrl+D now cleanly exits the cwd-change prompt during resume/fork flows instead of implicitly selecting an optionapp_metadata, branding, and labels), so clients can render more complete app cards without extra requestsprepare-commit-msg hook, with command_attribution override support (default label, custom label, or disable)remote_models feature flag to prevent fallback model metadata when it was disabled, improving model selection reliability and performancemodel/rerouted notification so clients can detect and render model reroute events explicitlymodel/list output to return full model data plus visibility metadata, avoiding unintended server-side filteringjs_repl stability issues, including reset hangs, in-flight tool-call races, and a view_image panic pathjs_repl) that can persist state across tool calls, with optional runtime path overrides/m_update, /m_drop) and expanded memory-read/metrics plumbingReadOnlyAccess policy shape for configurable read accessresponse.completed, and treated response.incomplete as an error pathlimit_name when merging partial rate-limit updates/statusline to configure which metadata appears in the TUI footer interactivelyrequirements.tomlrc filesview_image errors on text-only models and stripping unsupported image history during model switchesEnter sends immediately during running tasks while Tab explicitly queues follow-up inputresumeThread() argument ordering in the TypeScript SDK so resuming with local images no longer starts an unintended new session/debug-config slash command in the TUI to inspect effective configurationlog_dir so logs can be redirected (including via -c overrides) more easilythread/compact to the v2 app-server API as an async trigger RPC, so clients can start compaction immediately and track completion separatelycodex.rate_limits event, with websocket parity for ETag/reasoning metadata handlingunified_exec on all non-Windows platforms/debug-configEsc handling in the TUI request_user_input overlay: when notes are open, Esc now exits notes mode instead of interrupting the sessionrequest_rule guidance used in approval-policy prompting to correct rule behaviorcodex app <path> on macOS to launch Codex Desktop from the CLI, with automatic DMG download if it is missing~/.agents/skills (with ~/.codex/skills compatibility), plus app-server APIs/events to list and download public remote skills/plan now accepts inline prompt arguments and pasted images, and slash-command editing/highlighting in the TUI is more polishedCODEX_THREAD_ID, so scripts and skills can detect the active thread/session$PWD/.agents read-only like $PWD/.codexcodex exec hanging after interrupt in websocket/streaming flows; interrupted turns now shut down cleanlyrequestApproval IDs align with the corresponding command execution itemscf-ray and requestIdpersonality, and existing settings migrate forward.agents/skills, with clearer relative-path instructions and nested-folder markers supported/plan shortcut for quick mode switching/apps to browse connectors in TUI and $ insertion for app promptsupdated_at until the first turn actually startsthread/unarchive RPC to restore archived rollouts back into active sessionsconfig.toml, reducing the need to pass --scopes on each loginweb_search is now the default client behaviorweb_search tool now handles and displays all action types, and shows in-progress activity instead of appearing stuckcodex resume --last --json so prompts parse correctly without conflicting argument errorsrequest_user_input is now rejected outside Plan/Pair modes to prevent invalid tool calls--yolo now skips the git repository check instead of failing outside a repo/permissions command with a shorter approval set while keeping /approvals for compatibility/skill UI to enable or disable individual skillsthread/read and can filter archived threads in thread/listconfig.toml resolution and config/read can compute effective config from a given cwd~//...config.toml resolutionthreadId in both content and structuredContent, and returns a defined output schema for compatibilitySKILL.toml (names, descriptions, icons, brand color, default prompt) and surfaced in the app server and TUIspawn_agent accepts an agent role preset, and send_input can optionally interrupt a running agent before delivering the message/models metadata now includes upgrade migration markdown so clients can display richer guidance when suggesting model upgradesno_new_privs before applying sandbox rulescodex resume --last now respects the current working directory, with --all as an explicit overridecodex tool in codex mcp-server now includes the threadId in the response so it can be used with the codex-reply tool, fixing #3712. The documentation has been updated at https://developers.openai.com/codex/guides/agents-sdk/tuirequirement/list so clients can read requirements.toml and adjust agent-mode UX/elevate-sandbox commandLD_LIBRARY_PATH/DYLD_LIBRARY_PATH to avoid runtime issues. As explained in #8945, failure to pass along these environment variables to subprocesses that expect them (notably GPU-related ones), was causing 10x+ performance regressions! Special thanks to @johnzfitch for the detailed investigation and write-up in #8945/review <instructions> in TUI/TUI2 now launches the review flow instead of sending plain text/dev/null correctly to avoid misclassified changesthread/rollback so IDE clients can undo the last N turns without rewriting historyweb_search_cached to enable cached-only Web Search results as a safer alternative to live requestscodex exec resume[analytics] enabled=... config section to control analytics behaviorapply_patch parsing tolerant of whitespace-padded *** Begin Patch / *** End Patch markersCODEX_MANAGED_CONFIG_PATH from overriding managed config in production, closing a policy bypassCtrl+G to open the current prompt in your configured external editor ($VISUAL/$EDITOR) and sync edits back into the TUI.codex/config.toml, honor configurable project_root_markers, and merge that with system config like /etc/codex/config.tomltui2 transcript navigation with multi-click selection, a copy shortcut/affordance, and a draggable auto-hiding scrollbar/review compute diffs from the session’s working directory so base-branch detection works correctly with runtime cwd overridestui2 rendering/input edge cases (screen corruption, scroll stickiness, and selection/copy correctness)ripgrep download fails during CLI packaging so failures are easier to debugtui.scroll_* config settings (PR #8357)allowed_sandbox_modes to requirements.toml to constrain permitted sandbox modes (PR #8298)rmcp_client feature flag (PR #8087)codex-file-search (PR #8334)models.json) (PR #8168)/undo interacting destructively with git staging / ghost commits (PR #8303)contributing.md in experimental.md (PR #8311)/undo interacting destructively with git staging / ghost commits (PR #8303)contributing.md in experimental.md (PR #8311)AI 에이전트 플랫폼. CLI changelog는 없고, 큰 기능 출시와 커넥터 중심.
여러 Gmail·Google Calendar 계정을 Manus에 연결.
플래그십 에이전트 + 모바일 앱 개발.
대규모 병렬 리서치 — 수백 에이전트 동시 탐색.
클라우드 전용 머신에서 에이전트 실행.
AI 디자인 워크플로로 시안 제작.
어떤 브라우저든 AI 브라우저로 — 작업 자동화.
대면 회의 녹음 → AI 요약.
반복 작업 시간 기반 자동화.
유용한 대화를 프로젝트 지침으로.
프로젝트로 작업을 묶어 체계적으로 관리.
Slack에서 Manus 에이전트 직접 실행.
Manus 에이전트가 내 데스크톱을 직접 조작.
Manus Slides에 GPT Image 2 탑재 — AI 이미지 생성.
Manus로 만든 웹사이트를 한 번에 복사.
Manus 앱을 공유·테스트 후 Google Play·App Store 출시.
Manus 앱에 SEO 내장 — 검색엔진 노출 자동화.
스튜디오급 프레젠테이션 템플릿 — Nano Banana Pro.
Manus 앱에 커스텀 도메인 연결.
오픈 스탠더드 에이전트 스킬 통합 — 작업 범위 확장.
Manus 공식 학습 플랫폼 — 가이드·강의 제공.
맞춤 브라우저를 Manus 자동화 흐름에 연결.
팀 노하우를 재사용 가능한 에이전트 스킬로 저장.
Manus에 Stripe 연결 → 결제 처리부터 제품 출시까지.
Manus 커넥터로 Instagram 크리에이터 마켓플레이스 자동화.
Telegram에서 Manus 에이전트 직접 실행.
Manus에서 Similarweb으로 경쟁사 키워드·레퍼럴·페이지 분석.
작업 중 Manus가 적합한 커넥터를 자동으로 추천.
Google Drive 커넥터 업그레이드 — Workspace 전반 정밀 작업 가능.
Manus 프로젝트에 맞는 서버 유형을 호스팅 모드로 선택.
하나의 작업에서 여러 방향을 동시에 병렬 탐색.
빌드 결과물을 자동으로 지속 퍼블리시.
Manus에서 PowerPoint 호환 슬라이드 직접 생성·내보내기.
실행 전 계획을 먼저 수립 — 더 신중하고 정확한 작업 진행.
AI 도구 펀딩·인수·파트너십 동향.