ClawBox Starter

Your Personal AI Appliance

ClawBox Starter is the easiest way to run AI models locally. Plug it in, open the web dashboard, and start running Llama 3, Mistral, Phi, and 50+ other open-source models — no cloud APIs, no recurring fees, no data leaving your network.

Hardware Specifications

• AI-Optimized SoC with integrated NPU + GPU
• 16GB Unified Memory — run 7B-13B models natively
• 256GB NVMe Storage — store 5-8 models locally
• Gigabit Ethernet + Wi-Fi 6
• 15-25W Power — silent, fanless design
• Compact Form Factor — fits on any desk or shelf

Software Included

• Pre-configured AI runtime (llama.cpp optimized)
• Web management dashboard
• OpenAI-compatible REST API
• 1-click model marketplace (50+ models)
• Built-in AI router for multi-model setups
• OTA updates — always up to date

Performance

• Llama 3.1 8B: ~38 tokens/sec generation
• Mistral 7B: ~42 tokens/sec generation
• Phi-3 Mini: ~68 tokens/sec generation

Perfect For

Developers, hobbyists, small teams, and anyone who wants private, local AI without the complexity. Great for personal assistants, code completion, writing tools, and learning.

Open Source: 100% Apache 2.0 licensed. View source on GitHub.
Warranty: 2-year hardware warranty. 30-day returns for unopened units.
Shipping: Free US shipping. Ships within 5-7 business days.