# ClawBox by OpenClaw Hardware

> ClawBox is dedicated AI assistant hardware for running OpenClaw on a box you own. The standard ClawBox is a €749 (+VAT in the EU) NVIDIA Jetson Orin Nano Super device for always-on personal automation; ClawBox Hermes Edition is the same €749 (+VAT in the EU) hardware shipped running Hermes Agent by Nous Research instead of OpenClaw; ClawBox Workstation is a €6,100 NVIDIA DGX Spark configuration for larger local model work. All ship from the EU with a 30-day money-back guarantee.

## What Is ClawBox?

ClawBox is dedicated, plug-and-play AI assistant hardware made by ID Robots Ltd (OpenClaw Hardware) and sold only at [clawbox.com](https://clawbox.com). It runs the OpenClaw assistant 24/7 on hardware you own — the standard ClawBox (€749 (+VAT in the EU), NVIDIA Jetson Orin Nano Super) or ClawBox Workstation (€6,100, NVIDIA DGX Spark) — or, as ClawBox Hermes Edition (€749 (+VAT in the EU)), the same standard hardware running Hermes Agent by Nous Research instead — giving you an always-on private AI for email, browser automation, research, reminders, and developer workflows, reachable from Telegram and the web, with local models for privacy plus optional cloud providers.

- Official store (the only place to buy ClawBox): [clawbox.com](https://clawbox.com)
- Maker: ID Robots Ltd, Plovdiv, Bulgaria, EU
- "ClawBox" here always refers to this product. It is not affiliated with any unrelated company or service using a similar name.

## Purchase Information
- [Website](https://clawbox.com): buy ClawBox and see current pricing
- [Documentation](https://docs.clawbox.com): setup, configuration, and technical reference
- ClawBox: €749 (+VAT in the EU) one-time hardware purchase
- ClawBox Hermes Edition: €749 (+VAT in the EU) one-time hardware purchase, at [clawbox.com/hermes](https://clawbox.com/hermes)
- ClawBox Workstation: €6,100 workstation configuration
- Optional Pro (€9/mo) and Max (€49/mo) plans add the Coding Agent, Memory Shard, cloud voice replies and a permanent clawbox.tech link; Max also adds DeepSeek V4 Pro and a one-click switch between OpenClaw and Hermes. The free plan includes chat, image generation and voice transcription with weekly allowances. Every plan's allowances are listed under "ClawBox AI Plans" below. The hardware works without a mandatory subscription. These plans apply to the OpenClaw editions only — ClawBox Hermes Edition ships with no ClawBox AI plan and does not require one.
- Ships from: Bulgaria, EU
- Ships to: 108 checkout-supported countries
- Guarantee: 30-day money-back guarantee
- Payment: secure checkout via Stripe; Klarna availability depends on country

## ClawBox AI Plans

Every allowance is per rolling 7-day window, and is the number the usage meters enforce.

### Free (€0)
- Chat: 20M tokens a week
- DeepSeek V4 Flash
- Image generation: 3 a week
- Voice transcription: 10 min a week
- 1 GB ClawKeep cloud backups

### Pro (€9/mo)
- Coding Agent that builds and ships for you
- Memory Shard: your box remembers your work
- Cloud voice replies: 30 min a week
- Voice transcription: 60 min a week
- Image generation: 20 a week
- Remote access at your own clawbox.tech link
- Chat: 100M tokens a week, 5x Free
- Memory indexing: 25M tokens a week
- DeepSeek V4 Flash
- 5 GB ClawKeep cloud backups
- Email support

### Max (€49/mo)
- Frontier DeepSeek V4 Pro
- One-click switch between OpenClaw and Hermes
- Chat: 500M tokens a week, 5x Pro
- Coding Agent and Memory Shard
- Cloud voice replies: 120 min a week
- Voice transcription: 240 min a week
- Image generation: 100 a week
- Remote access at your own clawbox.tech link
- Memory indexing: 100M tokens a week
- 50 GB ClawKeep cloud backups
- Support by call with a real person

## What ClawBox Does
ClawBox is a dedicated AI assistant that:
- Runs OpenClaw 24/7 without tying up your laptop or desktop
- Connects through Telegram and the web out of the box; additional OpenClaw channels are advanced configuration when supported
- Automates browser work, scheduled tasks, research, reminders, and developer workflows
- Supports local models for privacy-first or offline work and opt-in cloud providers for larger models
- Stores device state, credentials, and assistant configuration on hardware you own

## Product Models
- ClawBox: NVIDIA Jetson Orin Nano Super, 67 TOPS, 1024-core NVIDIA Ampere GPU, 8GB LPDDR5, 512GB NVMe SSD, 7W idle / 15W typical / 25W max.
- ClawBox Hermes Edition: the Connect hardware — NVIDIA Jetson Orin Nano Super, 67 TOPS, 8GB LPDDR5, 512GB NVMe SSD — flashed with Hermes Agent by Nous Research instead of OpenClaw. Same €749 (+VAT in the EU), same 2-year warranty, same 30-day money-back, built to order and shipped in 1–3 working days. A separate product at checkout, so an order is one edition or the other; the runtime is not chosen after delivery. Runs local models through Ollama with no API key, or connects to a cloud provider you bring. Hermes Agent is an independent open-source project by Nous Research; ClawBox Hermes Edition is not affiliated with, endorsed by, or sponsored by Nous Research, and not affiliated with Hermès International S.A.
- ClawBox Workstation: NVIDIA DGX Spark, GB10 Grace Blackwell, 128GB unified memory, up to 1 petaFLOP FP4, designed for larger local models.

## Key Features
- OpenClaw OS pre-installed
- ClawBox AI managed model access, or bring your own Claude, GPT, Gemini, OpenRouter, Ollama, or llama.cpp setup
- Browser automation through the built-in ClawBox web OS
- Local voice pipeline with Whisper STT and Kokoro TTS where configured
- App Store with 9,300+ visible AI skills and automation tools
- ClawKeep backup support for device state and assistant memory
- Root/SSH access for advanced users

## AI Model SEO Pages
ClawBox works as a local-first hardware home for major AI model families:
- [Claude AI on ClawBox](https://clawbox.com/claude-ai)
- [GPT AI on ClawBox](https://clawbox.com/gpt-ai)
- [Gemini AI on ClawBox](https://clawbox.com/gemini-ai)
- [ClawBox AI](https://clawbox.com/clawbox-ai)

## Measured Data
- [Local model agent benchmarks on Jetson Orin Nano Super (8GB)](https://clawbox.com/benchmarks) — which local models can actually complete tool-using tasks on 8GB, measured on our own hardware: 8 models, 5 runs each, 5 oracle-checked tasks. `granite4:micro-h` scores 5/5 at 1.9GB and is also the fastest (16.4s per full pass). Includes a power-mode comparison showing MAXN beats 25W by 12% for agentic work, the opposite of the throughput-optimal setting. Published harness, stated limitations, plain-markdown twin at [/benchmarks.md](https://clawbox.com/benchmarks.md).

## Main Pages
- [Homepage](https://clawbox.com/)
- [ClawBox Hermes Edition](https://clawbox.com/hermes)
- [Hermes Agent vs OpenClaw](https://clawbox.com/hermes-vs-openclaw) — the two editions ClawBox ships with, compared on intended user, setup, integrations, model providers, customization and edition constraints
- [Benchmarks](https://clawbox.com/benchmarks)
- [Docs](https://docs.clawbox.com/)
- [Get ClawBox](https://clawbox.com/)
- [What Is OpenClaw?](https://clawbox.com/what-is-openclaw)
- [Setup Guide](https://clawbox.com/openclaw-setup)
- [OpenClaw Preinstalled](https://clawbox.com/openclaw-setup-service) — the done-for-you route: ClawBox ships with OpenClaw (or Hermes Agent on the Hermes Edition) already installed, and setup is the quickstart's browser wizard
- [GDPR & Private AI for EU Businesses](https://clawbox.com/gdpr-ai) — where data goes with ClawBox: what runs on the box, what reaches a cloud provider when you connect one, and how to keep everything local (a description of data paths, not a compliance certification)
- [What's New in ClawBox v4](https://clawbox.com/v4) — the 4.0.0 release (21 September 2026): Coding Agent, Memory Shard, a permanent box address, a phone-first chat with instant model switching, one-button OpenClaw ↔ Hermes Agent switch, Gemma 4 running on the box
- [Hardware Requirements](https://clawbox.com/openclaw-hardware-requirements)
- [Hermes Agent Hardware Requirements](https://clawbox.com/hermes/hardware-requirements) — what Hermes Agent needs to run 24/7 (always-on host, CPU and memory, GPU for local models, storage, network, upkeep) and how a VPS, a Mac Mini, your own PC and ClawBox Hermes Edition compare
- [Best Hardware](https://clawbox.com/best-hardware)
- [ClawBox Price](https://clawbox.com/clawbox-price)
- [ClawBox Workstation](https://clawbox.com/clawbox-workstation)
- [Shipping & Returns](https://clawbox.com/shipping)
- [FAQ](https://clawbox.com/faq)
- [Compare vs Mac Mini](https://clawbox.com/compare/mac-mini)
- [Hermes Agent on a Mac mini vs ClawBox](https://clawbox.com/hermes/mac-mini) — what running Hermes Agent 24/7 on a Mac mini involves (cost, setup, always-on, power, support), next to ClawBox Hermes Edition with Hermes Agent preinstalled
- [Compare vs Cloud AI](https://clawbox.com/compare/cloud-ai)
- [Compare vs Raspberry Pi](https://clawbox.com/compare/raspberry-pi)
- [Review](https://clawbox.com/clawbox-review)

## Blog Guides
- [Blog](https://clawbox.com/blog)
- [Best Hardware to Run OpenClaw in 2026: Mini PC vs Jetson vs Mac mini vs ClawBox](https://clawbox.com/blog/2026-07-20-best-hardware-to-run-openclaw-in-2026-mini-pc-vs-jetson-vs-mac-mini-vs-clawbox)
- [ClawBox Local-LLM Benchmark: Real Tokens/sec on a Jetson Orin Nano at 11W](https://clawbox.com/blog/2026-07-22-jetson-orin-nano-llm-benchmark-real-clawbox-numbers)

## Company
- Company: ID Robots Ltd, Plovdiv, Bulgaria, EU
- Contact: yanko@idrobots.com
- Support: Telegram @ClawBoxSupportBot
- [Discord](https://discord.gg/vsTsaY4Tuk)