OpenClaw Glossary

The core OpenClaw service that runs on your machine. It handles all connections to AI models, messaging platforms, and skills. Think of it as the central hub that orchestrates everything.

An AI assistant instance configured with specific capabilities, personality, and settings. You can have multiple agents with different roles (e.g., a research agent, a coding agent).

A temporary, specialized agent spawned by the main agent to handle specific tasks. Subagents work independently and report back when done. They're ephemeral — created for a task, then terminated.

An architecture where multiple specialized agents work together. For example, a router agent delegates tasks to research, coding, or creative agents based on the request.

A messaging platform connection (WhatsApp, Telegram, Slack, Discord, etc.). Channels let you communicate with OpenClaw from your preferred apps.

A conversation context between you and OpenClaw. Sessions maintain history and state. You can have multiple sessions across different channels.

OpenClaw's persistent storage of information about you, your preferences, and past interactions. Unlike ChatGPT, OpenClaw remembers things between conversations and across restarts.

The information sent to the AI model with each request. This includes the current conversation, relevant memory, and system prompts. Larger contexts use more tokens.

The text instruction you give to the AI. A well-crafted prompt leads to better responses. OpenClaw includes system prompts that define the agent's behavior.

A plugin that adds specific capabilities to OpenClaw. Skills can integrate with external services (weather, calendar), perform actions (file operations), or enhance AI abilities.

A pre-built automation or workflow that solves a specific problem. Recipes combine prompts, skills, and configurations into ready-to-use solutions like 'Morning Briefer' or 'Invoice Sorter'.

A scheduling system for automated tasks. Cron jobs run at specified times (e.g., every morning at 7 AM). Named after the Unix cron utility.

A periodic check-in that allows OpenClaw to be proactive. During heartbeats, the agent can check email, calendar, or other services and notify you of important updates.

The unit AI models use to measure text. Roughly 4 characters or 0.75 words per token. You're charged based on tokens used for both input and output.

A secret credential that authenticates you with AI providers like OpenAI or Anthropic. Keep it private! Never share your API key or commit it to public repositories.

A company that offers AI models via API. Common providers include OpenAI (GPT-4), Anthropic (Claude), Google (Gemini), and OpenRouter (multi-provider gateway).

The specific AI brain powering responses. Examples: GPT-4, Claude 3 Sonnet, Gemini Pro. Different models have different capabilities, speeds, and costs.

The directory where OpenClaw stores its configuration, memory, and files. Usually ~/.openclaw/ on your machine. You can view and edit files here directly.

An open-source tool for running AI models locally on your machine. With Ollama, you can use models like Llama 3 without paying for API calls — and with complete privacy.

A unified API gateway that gives access to 100+ AI models from various providers with a single API key. Great for trying different models or switching between them.

Running OpenClaw on your own computer or server instead of using a managed service. Self-hosting gives you complete control over your data and configuration.

The official managed hosting service for OpenClaw (coming soon). QuickBots handles servers, security, and updates — you just chat.