Home

ClawSandbox Wiki

Your personal AI workforce — running on the hardware you already own.

ClawSandbox lets you deploy and manage a fleet of isolated OpenClaw instances on a single machine. No cloud bills, no new hardware, no SaaS subscriptions — only the resources you already have.

If you have a Mac with Apple Silicon (or any Linux machine), you can spin up an entire OpenClaw fleet in minutes and connect each instance to the messaging platforms you already use.

What you get

• Multiple AI assistants running simultaneously, each with its own persona, skills, LLM provider, and messaging channel
• Full isolation — every instance runs in its own Docker container with a dedicated desktop
• Zero extra cost — the only ongoing expense is LLM API usage (and some providers offer free tiers)
• Web Dashboard — manage everything from your browser

10-minute quickstart roadmap

Step	What to do	Time
1	Install ClawSandbox	2 min
2	Launch the Dashboard	30 sec
3	Build the sandbox image	5 min (one-time)
4	Add an LLM provider	1 min
5	Create and configure an instance	1 min

Page index

Getting started

• Getting Started — prerequisites, install, first instance end-to-end

Dashboard

• Dashboard Guide — sidebar navigation, asset management, fleet management, image management
• Characters & Skills — persona system and skill management
• Soul Archive — save, browse, and clone instance souls

LLM Provider guides

• Anthropic (Claude) — Claude Sonnet, Haiku
• OpenAI (GPT) — GPT-4o, o3-mini
• Google (Gemini) — Gemini 2.0 Flash, Gemini 2.5 Pro
• DeepSeek — DeepSeek Chat, DeepSeek Reasoner

Messaging Channel guides

• Telegram
• Discord
• Slack
• Lark / Feishu

Reference

• CLI Reference — every command with examples
• FAQ — common issues and solutions