OpenClaw practical setup guide
OpenClaw is a self-hosted Gateway for AI agents. Run one Gateway on your own machine, connect your chat apps, and message your coding assistant from anywhere while keeping control of your setup.
Self-hosted Gateway
One local process owns sessions, routing, channels, memory, tools, and the browser Control UI.
Multi-channel by default
Use Telegram for the fastest start, then add WhatsApp, Slack, Discord, WebChat, Signal, Matrix, and more.
Agent-native
Built for coding agents with tools, sessions, skills, plugins, memory, multi-agent routing, and automation.
5-minute setup
Requirements: Node 24 recommended, Node 22.14+ supported, plus an API key from your model provider. Onboarding asks for the key and configures the Gateway.
curl -fsSL https://openclaw.ai/install.sh | bash openclaw onboard --install-daemon openclaw gateway status openclaw dashboard
iwr -useb https://openclaw.ai/install.ps1 | iex openclaw onboard --install-daemon openclaw gateway status openclaw dashboard
What should happen
- 1. Install script adds the OpenClaw CLI.
- 2. Onboarding chooses provider, API key, Gateway, and daemon/service setup.
- 3. `openclaw gateway status` should show the Gateway listening on port 18789.
- 4. `openclaw dashboard` opens the local Control UI.
Local dashboard default: http://127.0.0.1:18789/. Use Telegram first if you want the fastest phone-based channel setup.
Supported channels
Channels can run simultaneously. Text is broadly supported; media, reactions, groups, and pairing rules vary by channel.
Model providers
OpenClaw uses provider/model names. Start with onboarding, then tune defaults in config when you know your preferred model.