I get the same questions in my DMs every week. Here they all are, answered once and for all.
Is OpenClaw free?
Yes. OpenClaw itself is free and open source. You pay for the AI model API (Claude, GPT, etc.) and optionally for hosting. Typical cost: $20-30/month for personal use.
Can it run on my laptop?
Yes, but your agent dies when your laptop sleeps. For 24/7 operation, use a $5/month VPS or a Raspberry Pi.
Is it safe? Can it access my files?
It can access what you give it permission to access. By default, it runs in a sandboxed workspace. You explicitly grant access to external tools, APIs, and file paths.
How is it different from ChatGPT?
ChatGPT waits for you to talk to it. OpenClaw acts on its own — monitoring, automating, scheduling. It also has persistent memory across conversations.
What model should I use?
Claude Sonnet for most things. Opus for complex reasoning. GPT-4o as an alternative. The adaptive mode auto-selects and is what I recommend for beginners.
Can multiple people use the same agent?
Yes. Connect multiple channels (Discord server, Telegram group) and the agent serves everyone. Just be aware that conversation context is shared in group chats.
How do I update it?
npm update -g openclaw. That is it. Your config and memory files are preserved across updates.
Can it make phone calls?
Not natively. You could integrate with a telephony API like Twilio, but that is a custom integration. Out of the box, it handles text-based channels.
What happens if the API goes down?
The agent queues requests and retries when the API comes back. It does not lose messages. If you use multiple models, it can fall back to an alternative.
Is my data private?
Your data goes to whatever AI provider you configure (Anthropic, OpenAI, etc.). OpenClaw itself does not collect or transmit your data. Self-hosted means self-controlled.
Still have questions? The GitHub Discussions section is the best place to ask.