Where setup happens
Your server arrives with OpenClaw installed and its own web interface running. Everything your agent needs, the AI model and the chat channels, is configured inside that interface. OneClickClaw opens it for you the moment provisioning finishes, and the Open agent button on your dashboard opens it again any time.
Tip
You never paste keys into OneClickClaw. Your credentials go to your own server, where OpenClaw stores them.
Step 1: Pick your AI model
On the first visit OpenClaw shows its Model Setup page. Choose one of the sign-in methods: pair your ChatGPT or Codex subscription with a device code (open More sign-in options and press Pair next to ChatGPT Device Pairing), pair an xAI account, or paste an API key for OpenAI, Anthropic, xAI or another provider under Connect with an API key or token. OpenClaw sends one real request to the provider and marks the connection ready only after a model answered.
Warning
Skip the plain ChatGPT Login button: it sends you back to a localhost address on your own computer, which a hosted server cannot receive. ChatGPT Device Pairing is the one that works here.
Sign in with a provider
xAI OAuth [Pair]
More sign-in options: ChatGPT Device Pairing [Pair]
Connect with an API key or token
Provider: OpenAI, xAI, Anthropic, and more
[Connect & verify]
You can change the model later under Settings, then Models. That page lists every configured provider, the default model, the fallback chain and the thinking level.
Warning
If you add a provider such as OpenRouter or Google from the Add provider button on the Models page, that stores the key only. Pick the model you want in the Default model dropdown on the same page, and press Save. Until a default model is chosen the chat answers that it has no working inference.
Step 2: Connect a chat channel
Open Settings, then Channels. Each channel has a guided setup: WhatsApp shows a QR code to scan from your phone, Telegram and Discord ask for a bot token, and Slack, Signal, iMessage, Google Chat and Nostr follow their own short steps. Connected channels appear at the top of the page with a Running badge.
Connected channels
Telegram | Running
Add a channel
whatsapp [Set up]
telegram [Set up]
discord [Set up]
slack [Set up]
Note
Web chat works with no channel at all. The Home page of the interface is a full chat with your agent.
Step 3: Give your agent a name and a personality
Under Settings, then Agents, the Overview tab holds the display name, emoji and default model. The Files tab edits the workspace files that shape behaviour: SOUL.md for personality, USER.md for who the agent helps, AGENTS.md for working rules and MEMORY.md for long-term notes.
What your OneClickClaw dashboard shows
The dashboard keeps the server side: status, health, restarts, reinstalls, billing and the Open agent button. Until a model is configured, the overview shows a yellow banner. Our monitor checks your server every thirty seconds and clears the banner on its own once OpenClaw reports a model.
Warning
Hermes servers work differently: Hermes has its own dashboard, and the AI model for a Hermes agent is chosen from the AI Model tab on your OneClickClaw server page.