From today, you can run more than one OpenClaw bot on a single ClawCloud account. Each instance is fully independent — its own dedicated server, its own AI model, its own channel configuration, and its own API key.
What this makes possible
The most common use case: a personal Telegram bot and a separate Discord bot for a server or community you run. They don't share configuration, context, or credits.
Other setups that now work:
- Two Telegram bots on different accounts — one personal, one for a project
- A Claude-powered bot for long-form writing tasks and a GPT-4.1 bot for quick queries, side by side
- A managed-credits instance and a BYOK instance on the same account
- Separate OpenClaw bots for separate clients or workspaces
Each bot shows up independently in the dashboard with its own status indicator, credit bar, and model display.
How API keys work per instance
Each instance picks its own API mode during setup:
Managed mode — ClawCloud provisions a separate OpenRouter sub-key for that instance with its own monthly credit pool. Usage on one instance doesn't touch the credit budget of another. Each instance gets its own credit bar in the dashboard and resets independently on billing renewal.
BYOK — You supply an Anthropic, OpenAI, or Google API key directly for that instance. No credit tracking, no spending limits applied by ClawCloud.
The two modes can run at the same time across different instances. One managed bot and one BYOK bot on the same account is a supported setup.
Models are per-instance
Each bot has its own model selection. You can run Claude Sonnet on one instance and Gemini 2.5 Flash on another. Switching a model on one bot — via /model in chat or the deploy wizard — doesn't affect any other instance.
The full model catalog is available for each: curated models across Anthropic, OpenAI, Google, Meta, Mistral, and others. See OpenClaw API Costs Explained for a breakdown of per-model pricing if you're running multiple managed instances and want to keep costs predictable.
Provisioning a second bot
Go through the deploy wizard again from the dashboard. You'll pick a plan, channel, and API configuration the same way you did the first time. Once checkout completes, the new server provisions alongside your existing one.
Each instance is billed separately. If you're running one Lite and one Pro plan, those are two independent subscriptions.
What about self-managed alternatives
Tools like ZeroClaw and PicoClaw let you run OpenClaw on your own VPS. Multi-server setups there mean managing separate machines, separate daemon configs, and separate SSH access for each. That works, but it's manual.
Multi-instance ClawCloud is the managed version: no SSH, per-instance billing, dashboard monitoring, and automatic key rotation on billing reset — across as many instances as you need.
For a look at what OpenClaw can do across different channels and use cases, see the OpenClaw use cases section.