⚠️ 注意:この記事は2026年1月時点の「Clawdbot」のインストール記録です。 現在は「openclaw」に名称変更されており、パッケージ名やコマンド体系が異なる可能性があります。 最新の情報は openclaw公式リポジトリ を参照してください。
ただし、セットアップの流れや設計思想、注意すべきポイントは共通する部分が多いので、これから導入する人にも参考になるはずです。
Clawdbotを実機にインストールして、Gateway起動・TUI動作確認・Claude認証同期の仕組みまで確認した作業記録です。「動くかどうかわからない状態」から「証拠付きで動いた」まで、全コマンドと出力をそのまま載せています。

環境確認
まずはNodeのバージョンを調べます。
node -v
npm -v
結果は以下の通りです。
v24.12.0
11.8.0
Node v24という比較的新しいバージョンです。Clawdbotが動くかどうか、この時点ではわかりません。
Clawdbotのインストール
npm install -g clawdbot@latest
実行結果:
mamu@evo-x2:~/proxy$ npm install -g clawdbot@latest
npm warn deprecated gauge@4.0.4: This package is no longer supported.
npm warn deprecated are-we-there-yet@3.0.1: This package is no longer supported.
npm warn deprecated npmlog@6.0.2: This package is no longer supported.
npm warn deprecated tar@6.2.1: Old versions of tar are not supported, and contain
widely publicized security vulnerabilities, which have been fixed in the current
version. Please update. Support for old versions may be purchased (at exhorbitant
rates) by contacting i@izs.me
npm warn deprecated node-domexception@1.0.0: Use your platform's native DOMException instead
added 666 packages in 16s
124 packages are looking for funding
run `npm fund` for details
added 666 packages と出ているので、インストール自体は成功しています。出ているのは全部 deprecated 警告で、インストール失敗ではありません。
deprecated は「そのパッケージはメンテされていない/推奨されない」という注意喚起です。多くは依存の依存(Clawdbot本体が直接書いているとは限らない)で、この表示だけでは動作不能を示していません。tar@6.2.1 の文言は強めですが、npm側が「古いtarには脆弱性がある」と言っているだけで、修正は上流(依存チェーン側)の問題です。
インストール確認
clawdbot --version
which clawdbot
clawdbot help
実行結果:
mamu@evo-x2:~/proxy$ clawdbot --version which clawdbot clawdbot help
08:41:01 [agents/auth-profiles] synced anthropic credentials from claude cli
08:41:01 [agents/auth-profiles] synced openai-codex credentials from codex cli
2026.1.23-1
/home/mamu/.nvm/versions/node/v24.12.0/bin/clawdbot
🦞 Clawdbot 2026.1.23-1 (c8afa82)
It's not "failing," it's "discovering new ways to configure the same thing wrong."
Usage: clawdbot [options] [command]
Options:
-V, --version output the version number
--dev Dev profile: isolate state under ~/.clawdbot-dev,
default gateway port 19001, and shift derived ports (browser/canvas)
--profile Use a named profile
--no-color Disable ANSI colors
-h, --help display help for command
Commands:
setup Initialize ~/.clawdbot/clawdbot.json and the agent workspace
onboard Interactive wizard to set up the gateway, workspace, and skills
configure Interactive prompt to set up credentials, devices, and agent defaults
config Config helpers (get/set/unset)
doctor Health checks + quick fixes for the gateway and channels
dashboard Open the Control UI with your current token
reset Reset local config/state (keeps the CLI installed)
uninstall Uninstall the gateway service + local data (CLI remains)
message Send messages and channel actions
memory Memory search tools
agent Run an agent turn via the Gateway (use --local for embedded)
agents Manage isolated agents (workspaces + auth + routing)
acp Agent Control Protocol tools
gateway Gateway control / daemon / Gateway service (legacy alias)
logs Gateway logs
system System events, heartbeat, and presence
models Model configuration
approvals Exec approvals
nodes Node commands
devices Device pairing + token management
node Node control
sandbox Sandbox tools
tui Terminal UI
cron Cron scheduler
dns DNS helpers
docs Docs helpers
hooks Hooks tooling
webhooks Webhook helpers
pairing Pairing helpers
plugins Plugin management
channels Channel management
directory Directory commands
security Security helpers
skills Skills management
update CLI update helpers
status Show channel health and recent session recipients
health Fetch health from the running gateway
sessions List stored conversation sessions
browser Manage clawd's dedicated browser (Chrome/Chromium)
help display help for command
Examples:
clawdbot channels login --verbose
clawdbot message send --target +15555550123 --message "Hi" --json
clawdbot gateway --port 18789
clawdbot --dev gateway
clawdbot gateway --force
clawdbot agent --to +15555550123 --message "Run summary" --deliver
clawdbot message send --channel telegram --target @mychat --message "Hi"
Docs: docs.clawd.bot/cli
ここで注目すべき2つのポイント
① Node v24でもCLIは動いた。 clawdbot --version が通り、バイナリの場所も /home/mamu/.nvm/versions/node/v24.12.0/bin/clawdbot と確認できています。「Nodeが新しすぎて動かないかも」という懸念は、この時点では外れです。
② 既存の認証情報を自動検出している。 起動時のログに以下が出ました:
synced anthropic credentials from claude cli
synced openai-codex credentials from codex cli
ClawdbotがClaude CLIとCodex CLIの認証情報を自動検出して取り込んでいます。これが後に「APIキーを入力していないのに動く」仕組みの正体であることがわかります。
Onboarding(初期セットアップ)
clawdbot onboard
セキュリティ同意
🦞 Clawdbot 2026.1.23-1 (c8afa82) — I can run local, remote, or purely on vibes—results may vary with DNS.
┌ Clawdbot onboarding
│ ◇ Security ─────────────────────────────────────────────────────────────────╮
│ │ │
│ │ Please read: https://docs.clawd.bot/security │
│ │ │
│ │ Clawdbot agents can run commands, read/write files, and act through │
│ │ any tools you enable. They can only send messages on channels you │
│ │ configure (for example, an account you log in on this machine, or a │
│ │ bot account like Slack/Discord). │
│ │ │
│ │ If you're new to this, start with the sandbox and least privilege. │
│ │ It helps limit what an agent can do if it's tricked or makes a mistake. │
│ │ Learn more: https://docs.clawd.bot/sandboxing │
│ │ │
│ ◆ I understand this is powerful and inherently risky. Continue? │
│ ○ Yes / ● No │
└
ここが一番重要です。Clawdbotは「エージェント」として、コマンド実行、ファイル読み書き、有効化したツール経由の操作ができると宣言しています。強力なので最初にリスク同意を求めてきます。
今回はまず最小権限(least privilege)で始める方針で、Sandbox前提の動作確認をします。Yes で進めます。
QuickStart選択
◆ Onboarding mode
│ ● QuickStart (Configure details later via clawdbot configure.)
│ ○ Manual
まず「動く事実」を作ってから、あとで clawdbot configure で詰めればいいので、QuickStart を選択します。いきなり細かく設定すると沼りやすいので、動作確認→後で詰めるのが実践のコツです。
モデル/認証プロバイダ選択
◇ QuickStart ─────────────────────────╮
│ │
│ Gateway port: 18789 │
│ Gateway bind: Loopback (127.0.0.1) │
│ Gateway auth: Token (default) │
│ Tailscale exposure: Off │
│ Direct to chat channels. │
│ │
├──────────────────────────────────────╯
│ ◆ Model/auth provider
│ ● OpenAI (Codex OAuth + API key)
│ ○ Anthropic
│ ○ MiniMax
│ ○ Qwen
│ ○ Synthetic
│ ○ Google
│ ○ Copilot
│ ○ OpenRouter
│ ○ Vercel AI Gateway
│ ○ Moonshot AI
│ ○ Z.AI (GLM 4.7)
│ ○ OpenCode Zen
│ ○ Skip for now
ここでは Skip for now を選びました。理由は3つあります。
1つ目は、まずGateway起動・チャンネル連携・CLI操作までを確実に完了させたいこと。2つ目は、「脳みそ」は次の段階で無料寄り(Qwen/Ollamaなど)を検証できること。3つ目は、いきなり課金の話に入りたくないことです。
画面にある「OpenAI (Codex OAuth + API key)」は表記通りOAuth+APIキーが要件に含まれているので、「APIなしで完走できる保証」はありません。
モデルフィルタ
◇ Model/auth provider
│ Skip for now
│ ◆ Filter models by provider
│ ● All providers
│ ○ amazon-bedrock
│ ○ anthropic
│ ○ cerebras
│ ○ github-copilot
│ ○ google
│ ○ google-antigravity
│ ○ google-gemini-cli
│ ○ google-vertex
│ ○ groq
│ ○ minimax
│ ○ minimax-cn
│ ○ mistral
│ ○ openai
│ ○ openai-codex
│ ○ opencode
│ ○ openrouter
│ ○ vercel-ai-gateway
│ ○ xai
│ ○ zai
All providers のままにします。まだどのモデルにするか確定していないし、ここで特定プロバイダに絞ると不要な選択で迷うことになります。
デフォルトモデル
◇ Filter models by provider
│ All providers
│ ◆ Default model
│ ● Keep current (default: anthropic/claude-opus-4-5)
│ ○ Enter model manually
│ ○ amazon-bedrock/anthropic.claude-3-haiku-20240307-v1:0
│ ○ amazon-bedrock/anthropic.claude-3-5-haiku-20241022-v1:0
│ ...
Keep current を選びます。ここに表示されている anthropic/claude-opus-4-5 はデフォルト値というだけで、実際にモデルを呼び出す段階で認証がなければ動きません。ただ、ここで無理にモデルを変更しようとすると設定で詰まるので、まずはデフォルトのまま進めて、モデルはCLIで後から確実に差し替えます。
チャンネル選択
◇ Default model
│ Keep current (default: anthropic/claude-opus-4-5)
│ ◇ Channel status ────────────────────────────╮
│ │ │
│ │ Telegram: not configured │
│ │ WhatsApp: not configured │
│ │ Discord: not configured │
│ │ Slack: not configured │
│ │ Signal: not configured │
│ │ iMessage: not configured │
│ │ Nostr: install plugin to enable │
│ │ Microsoft Teams: install plugin to enable │
│ │ Mattermost: install plugin to enable │
│ │ Nextcloud Talk: install plugin to enable │
│ │ Matrix: install plugin to enable │
│ │ BlueBubbles: install plugin to enable │
│ │ Zalo: install plugin to enable │
│ │ Zalo Personal: install plugin to enable │
│ │ Tlon: install plugin to enable │
│ │ │
├─────────────────────────────────────────────╯
│ ◆ Select channel (QuickStart)
│ ● Telegram (Bot API) (not configured)
│ ○ WhatsApp (QR link)
│ ○ Discord (Bot API)
│ ○ Slack (Socket Mode)
│ ○ Signal (signal-cli)
│ ○ iMessage (imsg)
│ ○ Nostr (NIP-04 DMs)
│ ○ Microsoft Teams (Bot Framework)
│ ○ Mattermost (plugin)
│ ○ Nextcloud Talk (self-hosted)
│ ○ Matrix (plugin)
│ ○ BlueBubbles (macOS app)
│ ○ Zalo (Bot API)
│ ○ Zalo (Personal Account)
│ ○ Tlon (Urbit)
│ ○ Skip for now
Skip for now を選択。Bot作成・トークン発行・権限が絡むので、ここでやると長くなります。今回は「インストール〜onboard完走」が主役です。
Skills設定
◇ Select channel (QuickStart)
│ Skip for now
│ Updated ~/.clawdbot/clawdbot.json
│ Workspace OK: ~/clawd
│ Sessions OK: ~/.clawdbot/agents/main/sessions
│ ◇ Skills status ────────────╮
│ │ │
│ │ Eligible: 9 │
│ │ Missing requirements: 40 │
│ │ Blocked by allowlist: 0 │
│ │ │
├────────────────────────────╯
│ ◆ Configure skills now? (recommended)
│ ● Yes / ○ No
No を選択。「Missing requirements: 40」と出ている通り、ここから先は依存導入・許可設定の沼が待っています。Skillsは後からいつでもできるし、まずは本体が動くところまでを確定させます。
Hooks設定
◇ Configure skills now? (recommended)
│ No
│ ◇ Hooks ──────────────────────────────────────────────────────╮
│ │ │
│ │ Hooks let you automate actions when agent commands are issued.│
│ │ Example: Save session context to memory when you issue /new. │
│ │ │
│ │ Learn more: https://docs.clawd.bot/hooks │
│ │ │
├──────────────────────────────────────────────────────────────╯
│ ◆ Enable hooks?
│ ◻ Skip for now
│ ◻ 🚀 boot-md
│ ◻ 📝 command-logger
│ ◻ 💾 session-memory
Skip for now を選択。Hooksは便利ですが、最初から自動化を入れると挙動が読めなくなります。動作確認ができてから入れるのが安全です。
Systemdサービスのインストール
ここからは自動で進みます。
◇ Enable hooks?
│ Skip for now
│ ◇ Systemd ──────────────────────────────────────────────────────────────────╮
│ │ │
│ │ Linux installs use a systemd user service by default. Without lingering, │
│ │ systemd stops the user session on logout/idle and kills the Gateway. │
│ │ Enabling lingering now (may require sudo; writes /var/lib/systemd/linger). │
│ │ │
├──────────────────────────────────────────────────────────────────────────────╯
│ ◇ Systemd ─────────────────────────────╮
│ │ │
│ │ Enabled systemd lingering for mamu. │
│ │ │
├───────────────────────────────────────╯
│ ◇ Gateway service runtime ────────────────────────────────────────────╮
│ │ │
│ │ QuickStart uses Node for the Gateway service (stable + supported). │
│ │ │
├──────────────────────────────────────────────────────────────────────╯
│ ◐ Installing Gateway service…
│ Installed systemd service: /home/mamu/.config/systemd/user/clawdbot-gateway.service
│ ◇ Gateway service installed
systemd lingeringも自動で有効化されました。これでログアウトしてもGatewayが生き続けます。
Control UIとGateway起動確認
│ ◇ Agents: main (default)
│ Heartbeat interval: 30m (main)
│ Session store (main): /home/mamu/.clawdbot/agents/main/sessions/sessions.json (0 entries)
│ Missing Control UI assets. Build them with `pnpm ui:build` (auto-installs UI deps).
│ ◇ Control UI ─────────────────────────────────────────────────────────────────╮
│ │ │
│ │ Web UI: http://127.0.0.1:18789/ │
│ │ Web UI (with token): │
│ │ http://127.0.0.1:18789/?token=c2e496e1aaaa...aab9bc5 │
│ │ Gateway WS: ws://127.0.0.1:18789 │
│ │ Gateway: reachable │
│ │ Docs: https://docs.clawd.bot/web/control-ui │
│ │ │
├──────────────────────────────────────────────────────────────────────────────╯
⚠️ ここに表示されるtokenは共有鍵です。 そのまま他人に見せると誰でもControl UIに入れる可能性があります。clawdbot dashboard --no-open でいつでも再表示できるので、スクリーンショット等では必ず隠してください。
Onboard完了
│ ◆ How do you want to hatch your bot?
│ ● Hatch in TUI (recommended)
│ ○ Open the Web UI
│ ○ Do this later
Do this later を選択。TUIでの人格設定は次のステップに回して、今回は「インストール〜onboard完走」を確定させます。
◇ How do you want to hatch your bot?
│ Do this later
│ ◇ Later ───────────────────────────────────────────╮
│ │ │
│ │ When you're ready: clawdbot dashboard --no-open │
│ │ │
├───────────────────────────────────────────────────╯
│ ◇ Workspace backup ──────────────────────────────────────╮
│ │ │
│ │ Back up your agent workspace. │
│ │ Docs: https://docs.clawd.bot/concepts/agent-workspace │
│ │ │
├─────────────────────────────────────────────────────────╯
│ ◇ Web search (optional) ─────────────────────────────────────────────────────╮
│ │ │
│ │ If you want your agent to be able to search the web, you'll need an │
│ │ API key. Clawdbot uses Brave Search for the `web_search` tool. │
│ │ Set it up interactively: │
│ │ - Run: clawdbot configure --section web │
│ │ - Enable web_search and paste your Brave Search API key │
│ │ Alternative: set BRAVE_API_KEY in the Gateway environment. │
│ │ Docs: https://docs.clawd.bot/tools/web │
│ │ │
├─────────────────────────────────────────────────────────────────────────────╯
│ └ Onboarding complete. Use the tokenized dashboard link above to control Clawdbot.
Onboarding完走です。
Gatewayの稼働確認
clawdbot gateway status
実行結果:
🦞 Clawdbot 2026.1.23-1 (c8afa82)
I'm not saying your workflow is chaotic... I'm just bringing a linter and a helmet.
│ ◇ Service: systemd (enabled)
│ File logs: /tmp/clawdbot/clawdbot-2026-01-25.log
│ Command: /home/mamu/.nvm/versions/node/v24.12.0/bin/node
│ /home/mamu/.nvm/versions/node/v24.12.0/lib/node_modules/clawdbot/dist/entry.js
│ gateway --port 18789
│ Service file: ~/.config/systemd/user/clawdbot-gateway.service
│ Service env: CLAWDBOT_GATEWAY_PORT=18789
Service config looks out of date or non-standard.
Service config issue: Gateway service uses Node from a version manager;
it can break after upgrades.
(/home/mamu/.nvm/versions/node/v24.12.0/bin/node)
Service config issue: System Node 22+ not found;
install it before migrating away from version managers.
Recommendation: run "clawdbot doctor" (or "clawdbot doctor --repair").
Config (cli): ~/.clawdbot/clawdbot.json
Config (service): ~/.clawdbot/clawdbot.json
Gateway: bind=loopback (127.0.0.1), port=18789 (service args)
Probe target: ws://127.0.0.1:18789
Dashboard: http://127.0.0.1:18789/
Probe note: Loopback-only gateway; only local clients can connect.
Runtime: running (pid 71867, state active, sub running, last exit 0, reason 0)
RPC probe: ok
Listening: 127.0.0.1:18789
確認できた事実:
- systemd service enabled(常駐化できている)
- Gateway RPC probe: ok(疎通OK)
- Loopback-only (127.0.0.1)(ローカル専用で安全)
- Runtime: running / active(起動している)
運用上の警告
ただし、以下の警告も出ています:
Gateway service uses Node from a version manager; it can break after upgrades. System Node 22+ not found; install it before migrating away from version managers.
systemdサービスがnvm配下のNodeを直接参照しているため、nvmのバージョン切り替えやNode削除でサービスが壊れる可能性があります。対策は clawdbot doctor(または clawdbot doctor --repair)です。
systemdサービスの確認
systemctl --user status clawdbot-gateway --no-pager
注意:sudo はつけません。 systemctl --user はユーザーのsystemdセッションに繋ぎに行くので、sudo をつけるとDBUS/XDG系の環境変数が消えてユーザーのバスに繋げなくなります。これはClawdbotの問題ではなくLinuxの仕様です。
実行結果で確認できた事実:
- systemd user serviceで常駐している(
Loaded: ... enabled/Active: active (running)) - メモリ使用量は約374MB(ピーク1.4GB)
- Gatewayが 127.0.0.1:18789 で待ち受け(ローカル専用で安全)
- agent modelが
anthropic/claude-opus-4-5と表示されている
ヘルスチェック
clawdbot health
Gatewayの疎通確認です。Agent / Sessions の情報まで返してくるので、「骨格が動いている」証拠になります。
使用モデルの確認
clawdbot models list
実行結果:
🦞 Clawdbot 2026.1.23-1 (c8afa82)
— Your AI assistant, now without the $3,499 headset.
Model Input Ctx Local Auth Tags
anthropic/claude-opus-4-5 text+image 195k no yes default
ここで注目すべきは以下の3点です。
- Local = no:ローカル実行ではなく、クラウドのAnthropicを呼んでいる
- Auth = yes:Clawdbotがこのプロバイダの認証情報を持っている/設定済み
- Model = anthropic/claude-opus-4-5:デフォルトの「脳みそ」はClaude
つまり、APIキーを手入力していないのに動く状態ですが、「API不要」ではなく「認証が自動で通っている」のが実態です。
TUIで動作テスト
clawdbot tui
TUI(Terminal UI)を起動して、実際にエージェントに話しかけてみます。「こんにちは。今の状態を自己診断して」と入力した結果:
🔍 自己診断レポート
┌──────────────┬───────────────┬──────────────────────────────────────┐
│ 項目 │ 状態 │ 説明 │
├──────────────┼───────────────┼──────────────────────────────────────┤
│ IDENTITY.md │ ⚪ 未設定 │ 名前・性格などが空のテンプレート状態 │
│ MEMORY.md │ ❌ 存在しない │ 長期記憶ファイルがまだ作られていない │
│ memory/ │ ❌ 存在しない │ 日次ログ用ディレクトリもない │
│ USER.md │ ⚪ 未設定 │ あなたについての情報が空 │
│ BOOTSTRAP.md │ ✅ 存在 │ 初回起動用の指示書がまだある │
└──────────────┴───────────────┴──────────────────────────────────────┘
### 診断結果
🐣 生まれたてです。
これは完全に新しいワークスペースで、まだ誰が誰なのか、
何も決まっていません。BOOTSTRAP.mdが存在するということは、
これが初めての会話ということ。
### 技術的な状態
- Runtime: anthropic/claude-opus-4-5 on Linux
- Channel: webchat (メインセッション)
- Timezone: Asia/Tokyo
- Date: 2026-01-25
さて、BOOTSTRAPの指示に従うなら…
「僕は誰?あなたは誰?」 って聞くところから始めるんですが、どうします? 🤔
名前とか、どういうキャラで行くかとか、一緒に決めてもいいし、
あなたのことを教えてもらえると嬉しいです。
GatewayもTUIも動いています。エージェントがワークスペース(~/clawd)を読みに行って自己診断しました。「動くけど、人格と記憶が未設定」という初期テンプレート状態です。
ワークスペースの初期化
続けて「BOOTSTRAPの指示に従って、必要なファイルを作って初期化して。作ったファイル名と中身の概要を出して」と指示した結果:
✅ 初期化完了
┌───────────────────┬────────┬──────────────────────────────────────────────────┐
│ ファイル │ 状態 │ 概要 │
├───────────────────┼────────┼──────────────────────────────────────────────────┤
│ IDENTITY.md │ 🆕作成 │ 名前を「Kai(海)」に。デジタル・ファミリアとし │
│ │ │ て、落ち着きつつ好奇心旺盛なキャラ。絵文字は🌊 │
│ USER.md │ 📝更新 │ あなたの名前を「mamu」、タイムゾーンをAsia/Tokyo │
│ │ │ 、効率重視で技術的な理解があると記録 │
│ MEMORY.md │ 🆕作成 │ 長期記憶ファイル。今日の初期化と、mamuについて │
│ │ │ の初期印象を記載 │
│ memory/2026-01-25 │ 🆕作成 │ 今日の日次ログ。Day Oneとして初期化の記録 │
│ .md │ │ │
│ BOOTSTRAP.md │ 🗑️削除 │ 役目を終えたので削除 │
└───────────────────┴────────┴──────────────────────────────────────────────────┘
🌊 Kaiとして起動しました。よろしく、mamu。
何か手伝えることある?
AIがローカルにファイルを生成しました。ここで起きていることの構図はこうです:
- Claudeがやっていること:何を書くか(人格、説明文等)の判断と文章生成
- ローカルPCがやっていること:
~/clawd/配下へのファイル作成・更新・削除
つまり 「頭はClaude(クラウド)、手足はClawdbot(ローカル)」 です。
確認コマンド:
ls -la ~/clawd
sed -n '1,120p' ~/clawd/IDENTITY.md
sed -n '1,120p' ~/clawd/USER.md
「APIキーなしで動く」仕組みの正体
APIキーを手入力していないのにClaudeが動いている理由を、ログから確認します。
デフォルトモデルの確認
clawdbot models status --plain
anthropic/claude-opus-4-5
現在の「脳みそ」は anthropic/claude-opus-4-5 です。
認証同期の証拠
grep -n "synced\|refreshed credentials\|auth-profiles" /tmp/clawdbot/clawdbot-2026-01-25.log | tail -n 50
1: "synced anthropic credentials from claude cli"
profileId: "anthropic:claude-cli", type: "oauth", expires: "2026-01-25T09:39:08.953Z"
2: "synced openai-codex credentials from codex cli"
profileId: "openai-codex:codex-cli", expires: "2025-12-29T09:36:35.781Z"
1369: "wrote refreshed credentials to claude cli file"
expires: "2026-01-25T21:31:46.228Z"
これで仕組みが完全に判明しました。
① Claude CLIのOAuth認証をClawdbotが同期して使っている synced anthropic credentials from claude cli / type: "oauth" がその証拠です。APIキーではなく、Claude CLIのブラウザログイン(OAuth)の認証情報を流用しています。
② トークンの自動更新と書き戻しまでやっている wrote refreshed credentials to claude cli file は、ClawdbotがOAuthトークンを更新して、Claude CLI側の認証ファイルに反映していることを示しています。だから「気付いたら動き続ける」状態になります。
③ Codex CLIの認証も拾っている 将来OpenAI-Codex系モデルに切り替えても、同様に「キーを入れていないのに動く」可能性があります。
実際のAPI呼び出しログ
grep -nF 'embedded run start' /tmp/clawdbot/clawdbot-2026-01-25.log | tail -n 5
provider=anthropic model=claude-opus-4-5 thinking=low messageChannel=webchat
ClawdbotがAnthropicのClaude Opus 4.5を実際に叩いている確定的な証拠です。
まとめ
「APIキー入力してない」だけで、Claude CLIの認証を同期してクラウドを叩いている状態です。「課金ゼロ」とは別の話で、認証の出口が「APIキー課金」ではなく「Claude CLI/OAuth側の認証経路」になっているだけです。
もし完全にAPI課金ゼロにしたい場合は、Local = yes のモデル(Ollama等)を使う必要があります。
Gatewayの停止方法
systemd管理(通常はこれ)
systemctl --user stop clawdbot-gateway
systemctl --user status clawdbot-gateway --no-pager
次回起動も止めたいなら:
systemctl --user disable clawdbot-gateway
強制終了(最終手段)
pkill -f "clawdbot.*gateway" || true
この検証でわかったこと
- Node v24でもClawdbotは動く。 deprecated警告は出るが、CLIもGatewayもTUIも問題なく動作した。
- Onboardは「全部Skip」でも完走できる。 モデル・チャンネル・Skills・Hooksを全てスキップしても、Gateway起動まで到達する。後から
clawdbot configureで詰められる。 - Claude CLIの認証情報を自動同期する仕組みがある。 APIキーを手入力しなくても、Claude CLIにログインしていれば認証が通り、トークンの自動更新・書き戻しまでやってくれる。
- systemdがnvmのNodeを参照している問題がある。 運用としては
clawdbot doctor --repairで修正するか、システムNodeを別途インストールする必要がある。 - エージェントの「頭」と「手足」は分離している。 Claude(クラウド)が判断・生成を担当し、Clawdbot(ローカル)がファイル操作等を実行する構図。ローカルモデルに差し替えれば、完全にオフライン運用も可能な設計。
次にやること
clawdbot doctor --repairでsystemdのNode参照問題を修正する- ローカルモデル(Ollama/Qwen等)への「脳みそ交換」を検証する
- Telegram/Discord等のチャンネル連携を設定する
- Skillsの導入と運用テスト


