OpenClaw Glossary
Key terms and concepts for understanding the OpenClaw ecosystem, including components formerly known as Clawdbot and Moltbot.
OpenClaw
An open-source AI agent platform (formerly Clawdbot, then Moltbot) with 133K+ GitHub stars. It runs on your own infrastructure and connects to messaging channels like WhatsApp, Telegram, Slack, and Discord.
Clawdbot
The original name for the OpenClaw project. The project was later renamed to Moltbot, then to OpenClaw.
Moltbot
The second name for the OpenClaw project. It was renamed from Clawdbot and later became OpenClaw with a restructured codebase and new plugin system.
Gateway
The HTTP server component of OpenClaw that handles incoming requests, routes them to the AI backend, and enforces authentication and rate limiting.
ClawHub
OpenClaw's community marketplace with 700+ skills (plugins). Skills are community-submitted and range from web scrapers to calendar integrations. Not all skills are vetted for security.
skill.md
The standard file format for defining OpenClaw skills. It specifies what a skill does, its inputs/outputs, and execution logic. Skills extend what OpenClaw can do beyond basic conversation.
soul.md
Personality configuration file that defines how the AI speaks, what topics it avoids, brand guidelines, and behavioral guardrails. Each OpenClaw instance can have its own soul.md.
Multi-agent routing
An enterprise-level OpenClaw configuration where incoming messages are routed to different AI agents based on topic, user role, or channel. Used for scaling OpenClaw across departments or use cases.
Prompt injection
A security attack where malicious input tricks the AI into ignoring its instructions or performing unintended actions. OpenClaw deployments need specific defenses configured to mitigate this.
Model routing
A cost optimization technique where different LLM models are used for different types of queries — cheaper models for simple questions, more capable models for complex ones.
Token budget
A per-user or per-conversation spending limit that controls how many LLM tokens can be consumed. Used to prevent runaway API costs in OpenClaw deployments.
Compaction
The process of summarizing conversation history to reduce token usage while preserving context. Tuning compaction settings is a key cost optimization technique for OpenClaw.
Voice mode
An OpenClaw capability that enables speech-to-text and text-to-speech, allowing users to interact with the AI agent via voice through supported channels.
OpenClaw Experts
An independent marketplace that connects businesses with vetted OpenClaw specialists for setup, security, cost optimization, custom development, and enterprise deployment. Not affiliated with the OpenClaw open-source project.
Frequently Asked Questions
What are the key OpenClaw terms I should know?
The most important OpenClaw terms are: Gateway (the HTTP server), ClawHub (the skill marketplace), skill.md (skill definition format), and soul.md (personality configuration). Understanding these core components helps when setting up or customizing OpenClaw.
What is the difference between skill.md and soul.md?
skill.md defines what OpenClaw can do — it specifies a skill's inputs, outputs, and execution logic. soul.md defines how OpenClaw behaves — its personality, tone, topic restrictions, and guardrails.
What is ClawHub and is it safe?
ClawHub is OpenClaw's community marketplace with 700+ skills (plugins). Skills are community-submitted and not all are vetted for security. It is recommended to audit any ClawHub skill before installing it in a production deployment.