We just shipped ClawBox v2.2.3 Beta — the biggest update since launch. ClawBox is an AI hardware box built on the NVIDIA Jetson Orin Nano 8GB with a 512GB NVMe SSD, running OpenClaw (open source).
Here's what's new and why it matters for developers.
🖥️ Full Web OS
Open your browser, type your ClawBox IP, and you're looking at a real desktop environment. File manager, terminal, VS Code, installed apps — all in the browser. No SSH required.
This makes local AI accessible to people who aren't comfortable with the command line.
🏪 App Store
Browse and install AI skills with one click. Community-built skills with ratings, categories, and descriptions. Think VS Code extensions, but for your AI assistant.
⚡ MCP App Generation — The Killer Feature
This is the one that got us excited. Tell your AI:
"Build me a Social Media Tracker"
And it uses Model Context Protocol (MCP) to:
- Generate a full web application
- Create a desktop icon on your Web OS
- Launch the app
Single prompt → working desktop app. The AI scaffolds the frontend, connects to APIs, and deploys it locally.
🆓 ClawAI — Free AI Provider
Ships with a built-in free AI provider. No API keys, no accounts, no credit card. Select it during the 5-minute setup wizard and start chatting immediately.
You can always swap in your own API keys (Anthropic, OpenAI, etc.) or run local models.
🌐 Real Browser Automation
Full Chromium running on dedicated hardware. Websites can't tell it's automated — real browser, real hardware, real IP address.
Use cases:
- Web scraping without getting blocked
- Form filling and data entry
- Price monitoring
- Social media automation
- 24/7 background tasks
🌍 10-Language Setup Wizard
Plug in → pick your language → 5 steps → done in 5 minutes.
Tech Specs
- Hardware: NVIDIA Jetson Orin Nano 8GB + 512GB NVMe SSD
- AI Performance: 67 TOPS
- Local inference: ~15 tok/s for quantized 7B models
- Platforms: Telegram, WhatsApp, Discord, Web UI
- Software: OpenClaw (open source)
- Privacy: Everything runs locally, data never leaves the box
What's Next
We're working on expanding the App Store, improving MCP tooling, and adding more local model support.
If you want to try the Web OS or have questions about the MCP integration, drop a comment below.
📖 Full blog post with more details
🐣 Easter sale: use code EASTER10 for 10% off (ends Monday)





Top comments (0)