OpenClaw setup agent

기준일: 2026-07-26
공식 기준: OpenClaw setup agent

OpenClaw setup agent 문서는 OpenClaw 공식 문서(cli/openclaw)를 한국어로 정리한 가이드입니다. CLI reference and security model for the inference-backed OpenClaw setup and repair helper 명령·설정 키·코드 예시는 공식 문서를 그대로 보존하며, 해석과 절차 안내는 한국어로 제공합니다. 최종 동작은 설치된 CLI 버전과 공식 원문을 확인하세요.

핵심 요약

CLI reference and security model for the inference-backed OpenClaw setup and repair helper

한국어 가이드 범위: cli/openclaw 경로의 설정·명령·제약·예시를 학습용으로 재구성합니다.

문서 구성

공식 문서의 주요 섹션은 다음과 같습니다.

• 시작 시점
• What OpenClaw shows
• 예제
• Operations and approval
• Change history
• Switching to masked channel setup
• Setup bootstrap
• AI conversation
• CLI harness trust model
• Switching to an agent
• Message rescue mode
• 관련 문서

상세 내용

openclaw setup

OpenClaw ships with a built-in system agent — it speaks as "OpenClaw" — for local setup, repair, and configuration (formerly called Crestodian). It starts only after the effective default model completes a real turn. Fresh installs establish inference first; malformed config stays on the classic doctor path.

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

시작 시점

Running openclaw with no subcommand routes based on config state:

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

주요 항목:

• Config missing, or exists with no authored settings (empty, or only $schema/meta keys): starts guided onboarding with live AI verification.
• Config exists but fails validation: starts classic onboarding, which reports the issues and directs you to openclaw doctor.
• Config exists and is valid: opens the normal agent TUI. A reachable

What OpenClaw shows

Interactive OpenClaw opens the same TUI shell as openclaw tui, with an OpenClaw chat backend. The startup greeting covers:

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

주요 항목:

• config validity and the default agent
• the verified model OpenClaw is using
• Gateway reachability from the first startup probe
• the next recommended debug action

예제

openclaw
openclaw setup
openclaw setup --json
openclaw setup --message "models"
openclaw setup --message "validate config"
openclaw setup --message "setup workspace ~/Projects/work" --yes
openclaw setup --message "set default model openai/gpt-5.6" --yes
openclaw onboard --modern

status
health
doctor
validate config
setup
setup workspace ~/Projects/work
config set gateway.port 19001
config set-ref gateway.auth.token env OPENCLAW_GATEWAY_TOKEN
gateway status
restart gateway
agents
create agent work workspace ~/Projects/work
models
configure model provider
set default model openai/gpt-5.6
channels
channel info slack
connect slack
open channel wizard for slack
plugins list
plugins search slack
plugin install clawhub:openclaw-codex-app-server
talk to work agent
talk to agent for ~/Projects/work
audit
quit

Operations and approval

OpenClaw uses typed operations instead of editing config ad hoc.

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

Change history

The Ask OpenClaw page can show recent applied system-agent operations, Doctor migrations, Settings and CLI config writes, and manual edits to openclaw.json. The config journal detects external edits while the Gateway is watching, during an OpenClaw-owned write, or at the next startup after an offline edit.

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

Switching to masked channel setup

The local chat can hand control to the masked channel wizard:

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

open channel wizard for slack
channel info slack

Setup bootstrap

setup configures the remaining workspace and Gateway state after guided onboarding has already established inference. It writes only through typed config operations and asks for approval first.

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

setup
setup workspace ~/Projects/work

AI conversation

Interactive OpenClaw's free-form conversation runs through the same agent loop as regular OpenClaw agents, restricted to one ring-zero OpenClaw authority tool, openclaw, that wraps the typed operations. Read actions run freely, mutations require your conversational approval for that exact operation (see Operations and approval), and every applied write is audited and re-validated. The agent session persists, so OpenClaw has real multi-turn memory. If the verified inference route later stops working, return to openclaw onboard and repair it before continuing.

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

CLI harness trust model

Embedded runtimes and the Codex app-server harness enforce the ring-zero restriction directly: the run carries an OpenClaw tool allow-list with only the openclaw tool. For Codex, OpenClaw also disables environments, native execution, multi-agent, goal, app/plugin, skill/MCP, web-search, and request_user_input surfaces for that run. Codex still injects its inert native update_plan utility; it can update the model's temporary checklist but cannot write files or OpenClaw configuration. CLI harnesses do not consume OpenClaw's allow-list, so OpenClaw admits only backends whose own tool-selection contract can prove the same restriction:

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

주요 항목:

• Selectable backends, including Claude Code, launch with an empty native-tool
• Backends that declare no native tools receive the same dedicated OpenClaw
• Always-on or unknown native-tool backends fail closed before inference; they

Switching to an agent

Use a natural-language selector to leave OpenClaw and open the normal TUI:

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

talk to agent
talk to work agent
switch to main agent

/openclaw
/openclaw restart gateway

Message rescue mode

Message rescue mode is the message-channel entrypoint for OpenClaw: use it when your normal agent is dead but a trusted channel (for example WhatsApp) still receives commands.

위 내용은 공식 문서의 해당 섹션 요지입니다. 세부 플래그·기본값은 원문과 --help를 확인하세요.

주요 항목:

• Disabled when sandboxing is active for the agent/session; OpenClaw refuses remote rescue and points to local CLI repair.
• Default effective state is auto: allow remote rescue only in trusted YOLO operation, where the runtime already has unsandboxed local authority (tools.exec.security resolves to full and tools.exec.ask resolves to off, with sandbox mode off).
• Requires an explicit owner identity; no wildcard sender rules, open group policy, unauthenticated webhooks, or anonymous channels.
• Rescue is limited to owner DMs.
• Plugin search and list are read-only. Plugin install is always local-only (blocked in rescue, even when otherwise enabled) because it downloads executable code. Plugin uninstall is refused in both local OpenClaw and rescue; run openclaw plugins uninstall <id> from a terminal.
• Remote rescue cannot open the local TUI or switch into an interactive agent session; use local openclaw for agent handoff.
• Persistent writes still require approval, even in rescue mode.
• Pending approvals are one-use. Any newer rescue command for the same account, channel, and sender revokes the older plan; failed execution also consumes approval, so resend the command to retry.
• Every applied rescue operation is audited. Message-channel rescue records channel, account, sender, and source-address metadata; config-mutating operations also record config hashes before and after.
• Secrets are never echoed. SecretRef inspection reports availability, not values.
• If the Gateway is alive, rescue prefers Gateway typed operations; if it is dead, rescue uses only the minimal local repair surface that does not depend on the normal agent loop.

You, in a trusted owner DM: /openclaw status
OpenClaw: OpenClaw rescue mode. Gateway reachable: no. Config valid: no.
You: /openclaw restart gateway
OpenClaw: Plan: restart the Gateway. Reply /openclaw yes to apply.
You: /openclaw yes
OpenClaw: Applied. Audit entry written.

create agent work workspace ~/Projects/work model openai/gpt-5.6-sol
/openclaw create agent work workspace ~/Projects/work

pnpm test:docker:system-agent-rescue

pnpm test:live:system-agent-rescue-channel

관련 문서

주요 항목:

• CLI reference
• Doctor
• TUI
• Sandbox
• Security

실습 체크리스트

1. 공식 문서와 로컬 버전을 대조합니다: https://docs.openclaw.ai/cli/openclaw
2. 관련 CLI는 openclaw --help 및 하위 명령 --help로 옵션을 확인합니다.
3. 설정 변경 시 openclaw config / openclaw doctor로 유효성을 검사합니다.
4. Gateway·채널·플러그인 변경 후에는 필요 시 Gateway를 재시작합니다.

자주 쓰는 명령·설정 예시

openclaw
openclaw setup
openclaw setup --json
openclaw setup --message "models"
openclaw setup --message "validate config"
openclaw setup --message "setup workspace ~/Projects/work" --yes
openclaw setup --message "set default model openai/gpt-5.6" --yes
openclaw onboard --modern

status
health
doctor
validate config
setup
setup workspace ~/Projects/work
config set gateway.port 19001
config set-ref gateway.auth.token env OPENCLAW_GATEWAY_TOKEN
gateway status
restart gateway
agents
create agent work workspace ~/Projects/work
models
configure model provider
set default model openai/gpt-5.6
channels
channel info slack
connect slack
open channel wizard for slack
plugins list
plugins search slack
plugin install clawhub:openclaw-codex-app-server
talk to work agent
talk to agent for ~/Projects/work
audit
quit

open channel wizard for slack
channel info slack

setup
setup workspace ~/Projects/work

talk to agent
talk to work agent
switch to main agent

/openclaw
/openclaw restart gateway

관련 링크

• 공식 원문: cli/openclaw
• OpenClaw 문서 홈

이 가이드는 공식 문서를 한국어 학습용으로 재구성한 것입니다. 옵션 기본값·플래그 이름은 설치 버전에 따라 달라질 수 있습니다.