Autonomous.ai is shipping a consumer appliance that boots directly into OpenClaw: a 3D-printed pyramid case (styled after Las Vegas’s Luxor Hotel, complete with a light beacon) containing a Raspberry Pi 5 with 4 GB of RAM, a cooling fan, and a custom Linux stack. The device connects to users via Telegram, Slack, or Discord and runs AI agents that can read files, execute commands, and automate tasks on the local network, according to a review by New Atlas.
Reviewer Joe Salas reported a 10-minute setup time. He used the device to map unlabeled Ethernet cables across a new home, identify smart devices by IP and MAC address, manage a Frigate NVR security camera dashboard, and set up an SMB sandbox for file sharing.
Pricing and Token Economics
The original Intern includes 10 million free tokens per day through Autonomous’s ecosystem. Paid tiers run $20/month for 50 million daily tokens or $100/month for unlimited. Available LLM endpoints include Claude Haiku 4.5, Claude Opus 4.6, DeepSeek V4 Pro, MiniMax M2.7, Qwen 3.6 Plus, and GLM 5.1, along with an auto-selection option.
The $249 Intern 2 upgrades to an Orange Pi 4 Pro with 6 GB of LPDDR5 RAM, additional CPU cores, and a built-in 3-TOPS NPU. It adds voice capability and supports both OpenClaw and Hermes as agent runtimes. A developer edition with proper shell and SSH access is also available.
Cloud Exit in a Box
The device is tightly locked down by default, with no SSH access and communication routed through Autonomous’s app stack. Salas bypassed these restrictions within 20 minutes, bricking the device in the process and requiring a full re-flash from tech support. He then reconfigured the agent to use OpenRouter’s API instead of Autonomous’s token system, running on Meituan’s free LongCat-2.0 Preview model during its trial period and consuming roughly 500 million tokens in total experimentation.
The product positions local-first agent deployment as a consumer feature rather than an enterprise requirement. At sub-$250, it competes with traditional smart home hubs but with full agent orchestration capability. It pairs with the broader trend of enterprises and consumers moving agent workloads off hyperscaler clouds to cap costs and retain control.