Skip to main content

Glossary

This glossary covers the core concepts and technical terms used throughout the OpenClaw ecosystem, with translations in five languages to support the international community.

How to Use

Press Ctrl+F (or Cmd+F) to search for the term you need. Terms are sorted alphabetically by their English name.


Core Terms

A

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Agent代理 / Agent代理 / Agentエージェント에이전트An AI entity capable of autonomous action
API KeyAPI 金鑰API 密钥APIキーAPI 키Authentication key for accessing LLM services
Attack Surface攻擊面攻击面攻撃面공격 표면The range of a system exposed to attackers
Auth Token認證 Token认证 Token認証トークン인증 토큰Authentication credential for the Gateway API
Autonomous Agent自主代理自主代理自律型エージェント자율 에이전트An AI that can make decisions and act independently

B

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Bind Address綁定位址绑定地址バインドアドレス바인드 주소The network address the Gateway listens on
Browser Use瀏覽器操控浏览器操控ブラウザ操作브라우저 사용A skill that lets an agent control a web browser

C

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Channel頻道 / 通訊平台频道 / 通讯平台チャンネル채널A messaging platform the agent connects to (e.g., Telegram)
Channel Adapter頻道轉接器频道适配器チャンネルアダプター채널 어댑터Component handling a specific platform's protocol
ClawHavocClawHavoc 事件ClawHavoc 事件ClawHavoc事件ClawHavoc 사건The late-2025 malicious skill supply-chain attack
ClawHubClawHub 技能市集ClawHub 技能市场ClawHubスキルマーケットClawHub 스킬 마켓OpenClaw's skill-sharing marketplace
Compaction壓縮压缩コンパクション압축The process of compressing WAL entries into structured Markdown
Container容器容器コンテナ컨테이너Isolated environment for skill execution (Podman/Docker)
Context Window上下文窗口上下文窗口コンテキストウィンドウ컨텍스트 윈도우Maximum number of tokens an LLM can process at once
CVECVECVECVECVEPublicly disclosed security vulnerability identifier

D-E

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
DockerDockerDockerDockerDockerContainer platform (not recommended for OpenClaw)
Event Bus事件匯流排事件总线イベントバス이벤트 버스Internal communication mechanism between layers
Execution Layer執行層执行层実行レイヤー실행 레이어Layer 4: skills run in sandboxed containers

F-G

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Fallback回退 / 備用回退 / 备用フォールバック폴백Alternative when the primary LLM is unavailable
Function Calling函式呼叫函数调用ファンクションコーリング함수 호출Mechanism for an LLM to invoke external tools
Gateway閘道 / Gateway网关 / Gatewayゲートウェイ게이트웨이Layer 1: unified message ingestion entry point (port 18789)

I-L

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Intent Recognition意圖識別意图识别意図認識의도 인식The Reasoning Layer's process of determining user intent
LLM大型語言模型大型语言模型大規模言語モデル대규모 언어 모델The core reasoning engine of the AI
LLM RouterLLM 路由器LLM 路由器LLMルーターLLM 라우터Mechanism that routes requests to different models
Long-term Memory長期記憶长期记忆長期メモリ장기 기억Compacted, structured memory

M

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Manifest技能宣告檔技能清单文件マニフェスト매니페스트A skill's metadata and permission declarations
MCPMCPMCPMCPMCPModel Context Protocol — connects agents to external tools
Memory System記憶系統记忆系统メモリシステム메모리 시스템Layer 3: manages conversation history and user preferences
MoltyMolty (mascot)Molty (吉祥物)Molty (マスコット)Molty (마스코트)OpenClaw's lobster mascot
Multi-Agent多 Agent多 Agentマルチエージェント멀티 에이전트Architecture with multiple agents collaborating

O-P

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
OpenClawOpenClaw (養龍蝦)OpenClaw (养龙虾)OpenClawOpenClawOpen-source autonomous AI agent platform
Permission權限权限パーミッション권한Resources a skill is allowed to access
PodmanPodmanPodmanPodmanPodmanDaemonless container engine (recommended)
Prompt InjectionPrompt 注入Prompt 注入プロンプトインジェクション프롬프트 인젝션Attack that manipulates agent behavior via malicious instructions

R

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Rate Limiting速率限制速率限制レートリミット속도 제한Mechanism limiting the frequency of API requests
Reasoning Layer推理層推理层推論レイヤー추론 레이어Layer 2: handles intent recognition and response generation
RootlessRootlessRootlessルートレス루트리스Container execution mode that does not require root privileges

S

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
Sandbox沙箱沙箱サンドボックス샌드박스Isolated execution environment for skills
Sandbox Escape沙箱逃脫沙箱逃逸サンドボックスエスケープ샌드박스 탈출Attack technique that breaches container isolation
Skill技能技能スキル스킬An extensible function module the agent can execute
SOUL.mdSOUL.mdSOUL.mdSOUL.mdSOUL.mdMarkdown file defining agent personality and behavior
SSH TunnelSSH 通道SSH 隧道SSHトンネルSSH 터널Encrypted connection established via SSH
Supply Chain Attack供應鏈攻擊供应链攻击サプライチェーン攻撃공급망 공격Planting malicious code through third-party components

T

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
TokenTokenTokenトークン토큰The smallest text unit processed by an LLM
Tool Selection工具選擇工具选择ツール選択도구 선택The Reasoning Layer's process of choosing which skill to invoke
Tool Use工具使用工具使用ツール使用도구 사용Mechanism for an LLM to invoke external capabilities

V-W

EnglishTraditional ChineseSimplified ChineseJapaneseKoreanDescription
VirusTotalVirusTotalVirusTotalVirusTotalVirusTotalMulti-engine malware scanning platform
WALWAL (預寫日誌)WAL (预写日志)WAL (先行書き込みログ)WAL (선행 기록 로그)Write-Ahead Log — mechanism for real-time conversation recording
Warm Pool預熱池预热池ウォームプール워밍 풀Pre-created containers that reduce cold-start latency

Community Slang

TermDescription
Raising LobstersOpenClaw's nickname in Asian communities, from "Claw" (lobster claw)
MoltingA major OpenClaw version upgrade (from the Molty mascot's molting imagery)
ShowcaseA project demo shared on Reddit
Day 1An achievement completed on the first day of installing OpenClaw
nukeCompletely resetting OpenClaw (deleting all configs and memory)
skill shoppingBrowsing and selecting skills on ClawHub

Abbreviations Reference

AbbreviationFull FormDescription
APIApplication Programming InterfaceInterface for software-to-software communication
CLICommand Line InterfaceText-based interface for interacting with software
CVECommon Vulnerabilities and ExposuresPublic vulnerability identification system
CVSSCommon Vulnerability Scoring SystemStandardized vulnerability severity scoring
DNSDomain Name SystemTranslates domain names to IP addresses
LLMLarge Language ModelAI model trained on large text datasets
MCPModel Context ProtocolProtocol connecting agents to external tools
NVRNetwork Video RecorderDevice for recording video streams
OAuthOpen AuthorizationOpen standard for token-based authorization
OCIOpen Container InitiativeStandards for container formats and runtimes
RCERemote Code ExecutionAbility to run code on a remote machine
RESTRepresentational State TransferArchitectural style for web APIs
RSSReally Simple SyndicationWeb feed format for content distribution
SSEServer-Sent EventsOne-way server-to-client push mechanism
SSHSecure ShellEncrypted protocol for remote access
STTSpeech to TextConverting audio to written text
TLSTransport Layer SecurityEncryption protocol for network communication
TTSText to SpeechConverting written text to audio
VPNVirtual Private NetworkEncrypted network tunnel
WALWrite-Ahead LogLogging mechanism that writes before committing
WSLWindows Subsystem for LinuxLinux compatibility layer on Windows

Version Naming Convention

OpenClaw follows Semantic Versioning:

v3.2.1
│ │ │
│ │ └── Patch: Bug fixes, security updates
│ └──── Minor: New features, backward-compatible
└────── Major: Breaking changes

Further Reading