Personal computers were built for people to operate. Otto One is built so nobody has to. The New York startup Otto launched it on November 10 as a compact, screenless machine whose only job is to run an AI agent around the clock. The company says it had already sold 15,000 units before the formal introduction.
What you get
There’s no keyboard, mouse or monitor. You plug it in, pair it with the Otto app and talk to your agent from a phone or desktop. The agent keeps working when you close your laptop or walk away. A reviewer found a small, passively cooled Intel N95 box running Otto OS, the company’s software layer built on Debian Linux.
Pricing is $599 for 8GB of memory and 256GB of storage, described as introductory, or $799 for 16GB and 512GB.
Privacy is the core of the pitch. The agent’s memory, chat history, automations and actions stay on the device, not on Otto’s servers. Traffic between your phone and the box is end-to-end encrypted with the same protocol family Signal uses, and Otto’s servers only relay ciphertext. API keys you add never leave the device, and unpairing wipes the owner’s data.
The software is the product
CTO Quinn Finney’s point is that everyone is building agents that can act, but few are building the part that decides what an agent may do, what it may spend and what needs the owner’s sign-off. Otto OS is meant to be that control layer, with the hardware as the first way in.
Otto isn’t alone. AMD has started promoting agent computers as a category, and other local AI boxes are aimed at the same buyer. The argument for all of them is the same: an always-on agent needs somewhere to live, and a dedicated box at home avoids handing your files and accounts to someone else’s cloud. The open question is how many people want an agent busy enough to need its own machine.