Zo Computer — Your Personal Cloud Computer

Related guides: Personal OS · Git · The Terminal

Preface

Your laptop is a local computer: powerful, private, and asleep in your backpack half the day. A Zo Computer (zo.computer) is your personal cloud computer: a real Linux server that stays on 24/7, paired with an AI agent that knows how to operate it. You reach it from anywhere — web chat, text message, email — and it keeps working after you close your laptop.

This is not “another AI product.” It’s the second computer in the two-computer model from Class 1: a machine that is always available, persistent, and on the internet. Historically, getting your own server meant a rite of passage through Linux, SSH, DNS, ports, and deployment configs. Those technologies still exist — Zo’s agent just knows how to operate them on your behalf. You still need to know what’s running, who can reach it, and how to stop it.

By the end of this guide you should be able to:

Basics

1. Why use Zo? What a personal cloud computer does for you

Everything below follows from three properties: always on, persistent, connected to the internet.

The two-computer model in one line: local for private, interactive, hands-on work; cloud for anything that must be available, persistent, or public. Reach for Zo because the task needs its properties, not because “cloud” sounds fancier.

2. Powering Zo with your Codex or Claude Code subscription

Zo’s agent needs a frontier model behind it. By default it uses Zo’s built-in AI credits — but the better move for this course is to connect the subscription you already have. Zo officially supports running the real first-party CLIs on your Zo box, logged into your account, and using them as Zo’s model provider “instead of Zo’s built-in credits.”

Codex / ChatGPT (Plus or Pro) — recommended:

  1. In ChatGPT (on any device): Settings → Security → “Enable device code authorization for Codex.”
  2. In Zo’s terminal: run codex and choose “Sign in with Device Code.”
  3. Open the shown URL on your phone or laptop, enter the one-time code. Done.
  4. In Zo: Settings → AI Providers, select Codex as the model behind Zo’s chat.

Claude Code (Pro or Max):

  1. In Zo’s terminal: run claude. It prints a login URL — the URL may wrap across several terminal lines; join it into one line before opening it in your browser.
  2. Complete the login on your own device, paste the code back.
  3. Select Claude Code under Settings → AI Providers.

Why this matters: your flat-rate subscription now powers an always-on agent, instead of paying per-token for the same model twice. One honest caveat: OpenAI explicitly permits subscription sign-in from external tools, while Anthropic’s terms are more restrictive about automated third-party use of Claude subscriptions — so for scheduled/background workloads, the Codex path is the cleaner choice.

(Full docs: docs.zocomputer.com/codex and docs.zocomputer.com/claude-code.)

3. The most powerful features of Zo

Pricing shape (for context): a free tier exists but is not always-on; paid tiers (from ~$18/mo “Basic”: 24/7 uptime, 4 cores / 32GB RAM, monthly AI credits) are what make Zo an actual server. Always-on is the point — budget for a paid tier plus your model subscription.

4. Syncing your Personal OS onto your Zo

Your Zo becomes dramatically more useful once it holds a copy of your Personal OS — then the agent you text from your phone has the same map, plans, and projects as the agent on your laptop.

Two ways, in order of preference:

Git + GitHub (recommended, durable). Zo comes with the GitHub CLI (gh) pre-installed. In Zo’s terminal (or just ask your Zo agent to do this):

gh auth login        # Zo's docs use a GitHub personal access token here;
                     # ask your Zo agent to walk you through creating one
cd /home/workspace
gh repo clone <your-username>/personal-os

From then on it’s the standard sync loop: pull before working, commit and push after. Tell your Zo agent “pull my personal OS before you start” — or better, make it a standing instruction in your AGENTS.md.

Zo’s built-in file sync (quick start). Zo can mirror a folder from your computer without any Git knowledge — fine for week one. Its limitation is that it’s a mirror, not a history: no commit log, no conflict resolution, no third machines. Graduate to Git when your Personal OS starts holding real work.

Once synced, a powerful pattern opens up — agents collaborating through the repo. Some ways to get your systems talking to each other, roughly in order of directness:

  1. MCP: connect ChatGPT/Claude/your local agent to Zo’s MCP server, and it controls Zo directly (files, shell, messages, schedules).
  2. The shared repo as a relay: one agent (say, ChatGPT on your phone) writes a plan and pushes it to your personal-os repo; your Zo agent — on a schedule or on request — pulls the repo, finds the plan, and executes it. No special integration needed; the Personal OS is the interface.
  3. SSH: your local terminal (and local agents) operate Zo directly — the Advanced section below.

Advanced

5. SSH into your Zo from your own terminal

SSH gives you (and your local agents) a direct terminal on your Zo — no browser tab needed. Setup is one-time (full docs; your Zo agent can do most of it for you):

  1. On your laptop, create a key pair if you don’t have one: 
    ssh-keygen -t ed25519 -C "you@example.com"
cat ~/.ssh/id_ed25519.pub        # copy this line — it's the *public* key, safe to share
  2. In Zo’s terminal, add that public key to ~/.ssh/authorized_keys.
  3. In Zo’s UI, create the SSH service: Sites → Services → add service — Label ssh, Local port 2222, Type tcp, Entrypoint /usr/sbin/sshd -D -p 2222. Zo assigns you a public host and port (e.g. ts1.zocomputer.io:10872 — yours will differ).
  4. From your laptop: 
    ssh -p <your-port> root@<your-host>
  5. Make it one word by adding to ~/.ssh/config on your laptop: 
    Host zo
  HostName ts1.zocomputer.io
  Port <your-port>
  User root
  IdentityFile ~/.ssh/id_ed25519
    Now it’s just ssh zo — and in cmux, cmux ssh zo puts your cloud computer in the sidebar next to your local workspaces.

Bonus unlocked by the same setup: point Cursor or VS Code’s Remote-SSH at zo and open /home/workspace — your cloud computer as a full remote development environment.

(Direction check: this section is you SSH-ing into Zo. Zo also documents the reverse — letting Zo control your computer over SSH — which is a different, more sensitive grant; see docs.zocomputer.com/ssh-computer.)

6. Running Codex and Claude Code on your Zo

You already logged the CLIs in once in section 2 (to power Zo’s chat). The same installs mean you can run them interactively on the box — the full cloud-agent workflow:

ssh zo
cd /home/workspace/personal-os
codex          # or: claude

What you get: a frontier agent with your subscription, standing in your Personal OS, on a machine that never sleeps. Things to know:

Practice

  1. Send your Zo a text and an email; have it reply. Save its contact info in your phone.
  2. Connect your subscription: complete the Codex (or Claude Code) login and switch Zo’s AI provider to it. Confirm with a chat message that it’s using your subscription.
  3. Clone your Personal OS onto Zo with gh, then prove the loop: edit on the laptop → push → ask your Zo agent (by text!) to pull and summarize what changed.
  4. Create one real automation in plain English — a morning briefing, a page watcher — and let it run for a few days. Inspect a saved run.
  5. Host something tiny and share the URL with a friend: a page on your Space, or a small app as a Service. Then practice the off-switch: find it and stop it.
  6. (Advanced) Complete the SSH setup, ssh zo from cmux, run an agent inside your Personal OS on Zo, and have it commit-and-push its work.
  7. (Advanced) Add Zo’s MCP server to your local Claude Code or Codex and ask your local agent to list the files on your cloud computer.

Proof of understanding

Further reading