Question 1

Why do I need an otto device to run agents?

Accepted Answer

Tools like ChatGPT and Claude are great for back-and-forth conversations, but they're not built to act as long-lived workers — they don't run in the background, hold a role, or take action while you're asleep. Otto runs OpenClaw, the leading open-source framework for always-on agents that actually do. The catch is OpenClaw needs a computer to run on. You can run it on your laptop, but your agent dies when you close the lid. You can spin up a cloud VM, but cloud IPs get blocked on most of the sites that matter — LinkedIn, Reddit, Instagram — because they look like bots. A dedicated device on your home network, always on, browsing like a real person, is the right answer. Otto makes that setup take five minutes instead of a weekend.

Question 2

What do I need to get started?

Accepted Answer

An otto device and the otto app on your phone. The device ships with otto OS pre-installed — Runtime, Gateway, and Passkey all configured out of the box. Plug it in, open the app, tap to pair over NFC. You can have your first agent running in under five minutes. No terminal required.

Question 3

How is this different from just buying a Mac Mini?

Accepted Answer

You can use an otto device like any other Linux machine if you want — nothing is locked down. The difference is what you get out of the box. A Mac Mini ships as a general-purpose desktop: you're installing OpenClaw yourself, hardening the security config, wiring up CLIs, figuring out how to manage it headlessly. Otto ships purpose-built: OpenClaw pre-installed and pre-configured, NFC pairing, a proper management app, and otto OS tuned specifically for agent workloads — all at a lower price. If you want to do the Mac Mini setup, you can do it on an otto device too. You'd just be ignoring everything we already did for you.

Question 4

Do I need to know how to code?

Accepted Answer

No. The app lets you deploy agents and configure them with plain language. If you want to build custom agents, fork the runtime, or SSH into the device directly, you can — but none of that is required to get started.

Question 5

Can I use an OpenClaw agent I already set up?

Accepted Answer

Yes. Upload your existing config through the app and it runs as-is. You can also build new agents from scratch in otto OS without touching a terminal — either way works.

Question 6

What's the difference between Otto One and Otto Pro?

Accepted Answer

Otto One is the entry device — Intel N97, 8GB DDR4, 256GB SSD, $599. It handles local storage, credentials, and task execution. Inference runs in the cloud, which means lower cost and faster models. Otto Pro adds local inference hardware so your agents can run open-weight models on-device. Otto Pro is coming soon, starting at $999.

Question 7

What's in the box?

Accepted Answer

The otto device, a power adapter, and a short ethernet cable. No keyboard, no monitor — it's a headless agent computer by design.

Question 8

Is it loud? Does it have a fan?

Accepted Answer

Otto One is fanless and completely silent. It's designed for continuous, unattended operation — you can leave it running 24/7 without thinking about it.

Question 9

Does it need to be connected to the internet?

Accepted Answer

For cloud model routing, external tool access, or the relay that connects to the mobile app, yes. For fully local agent tasks on Otto Pro with local inference, you can run entirely offline.

Question 10

How many agents can it run at once?

Accepted Answer

A single otto device is well-suited for a small team's worth of agents — think exec assistant, research, ops, finance monitoring running concurrently. 5–10 agents is a reasonable target depending on task complexity and which models they're using. You can connect multiple devices under one account and manage them together from the same app.

Question 11

What's OpenClaw? Is the Runtime open source?

Accepted Answer

OpenClaw is an open-source agent harness that otto Runtime is built on. It's widely used, well-tested, and handles everything a runtime harness needs. Because it's open source, we can read it, understand it, and optimize on top of it. Otto's runtime itself — including our tuned defaults, agentic loop behavior, and otto-specific optimizations — is not open source, though we stay compatible with the OpenClaw ecosystem.

Question 12

Which models does Gateway support?

Accepted Answer

Gateway supports major cloud providers — Anthropic, OpenAI, Google — and local open-weight models on Otto Pro hardware. We're adding providers continuously. Local models run on-device with no data leaving your network. Cloud models are routed through Gateway with a single bill.

Question 13

How does Passkey store my credentials?

Accepted Answer

Credentials are encrypted and stored on your otto device — never on otto servers. When an agent needs access to a service, it requests a short-lived token from the device. Your raw credentials are never passed to the agent or the model, and they can't be exposed in a cloud breach because they're never in our cloud.

Question 14

Do I have to use the app?

Accepted Answer

Not at all. Otto is a Linux box running OpenClaw — if you want to SSH in, use the OpenClaw CLI directly, and manage everything yourself, that works exactly as it would on any other machine. The app is a service layer built to make the 90% case easier. Everything in the app is OpenClaw underneath.

Question 15

Does my data ever leave my network?

Accepted Answer

Your files, credentials, and browser sessions stay on the device and never touch otto servers. The outbound traffic is whatever your agents do themselves — API calls, web browsing, emails sent — same as if you were doing it manually. Because agents run from a real local browser on your network, they're not identifiable as cloud bots, which matters more than most people expect.

Question 16

What platforms is the app on?

Accepted Answer

iOS, Android, and Desktop. The app is your primary interface for everything — chat with your agent, approve tool calls, manage Passkey connections, view spend, and monitor agent activity.

Question 17

How is it different from the Claude or ChatGPT app?

Accepted Answer

Those apps connect you to a cloud model hosted by someone else. The otto app connects you to your own agent running on your own hardware. You get chain-of-thought visibility, tool call approvals, push notifications for human-in-the-loop moments, and control over all otto OS features — Passkey, Gateway, Identity, Spend. It's a control surface, not just a chat window.

Question 18

How does the relay work when I'm away from home?

Accepted Answer

The app connects to your device through otto's encrypted relay — end-to-end encrypted, routed only through otto infrastructure. No ports to open, no VPN to configure. It just works.

Question 19

Will I get spammed with notifications?

Accepted Answer

Notifications are opt-in per agent and per event type. By default you get a summary when a task completes. You can configure this per agent in the app.

Question 20

When will my Otto ship?

Accepted Answer

Orders placed on or after August 12, 2026 are estimated to ship in January 2027. This date is an estimate and may shift — we'll email you before your unit ships so you can confirm your address and final total.

Question 21

What does my purchase include?

Accepted Answer

The Otto One hardware, forever. A $20 fully refundable deposit reserves your unit. We'll email before shipment so you have 7 days to confirm your address and final total before the remaining balance is charged. Shipping is a flat $15 per unit worldwide. No subscription, no strings — the hardware is yours.

Question 22

How am I billed for cloud model usage?

Accepted Answer

Cloud model costs are routed through Gateway. We take a small transaction fee on cloud spend to cover our own infrastructure and hedging costs — you pay for what you use, not a monthly seat fee. Local inference on Otto Pro hardware runs free through Gateway.

Question 23

Is Gateway a subscription?

Accepted Answer

No. Gateway is part of otto OS, not a separate SaaS product. You were going to spend on tokens anyway — we sit at the OS layer and help you manage that spend better, taking a small transaction fee on cloud usage only.

Question 24

What's the pricing for Identity?

Accepted Answer

Identity (managed email, phone, and SMS for your agents) is $5.99 per agent per month. Full pricing will be confirmed closer to launch.

Question 25

When does Spend ship?

Accepted Answer

Spend (virtual agent cards and accounts) is in development. Order a device to get early access when it launches.

Frequently asked questions

Looking for how-to guides?