DOCUMENTATION

Openclaw
technical reference

Practical guides for deploying, configuring, and extending Openclaw AI Agent, from initial setup to production-grade multi-agent workflows.

Browse by Topic

Getting Started

Install Openclaw, run the onboarding wizard, configure your first AI model, and verify the runtime is healthy.

Gateway & Configuration

Manage the always-on gateway process, configure authentication, model failover, and the full openclaw.json reference.

Agent Playbooks

Concrete task automation patterns — cron jobs, event hooks, sub-agents, and real-world workflow implementations.

Channel Integrations

Connect Openclaw to Telegram, Discord, WhatsApp, Teams, Matrix, and other platforms. Covers auth, group rules, and message formatting.

Memory, Skills & Tools

How agent memory works, the SOUL.md personality file, ClawHub skill marketplace, browser control, shell exec, and web fetch tools.

Security & Troubleshooting

Docker sandboxing, filesystem access control, secret management, prompt injection mitigation, and the 60-second diagnostic runbook.

How Openclaw Works

Openclaw runs a persistent gateway process that routes messages from connected channels to a configured AI model. Each conversation is backed by a Markdown-based memory workspace. Agents can execute shell commands, call external APIs, spawn sub-agents, and interact with browsers — all scoped within configurable sandboxes. This documentation covers every layer: runtime setup, model configuration, channel wiring, skill authoring, and production hardening.