Connect your first VPS in 30 seconds

Ohuriya AI is an AI DevOps Copilot for VPS owners—you do not need to memorize ssh, systemctl, and journalctl on day one. You need a server, an account, and one install line. The install itself takes about 30 seconds. The rest is chat—and you stay in control because nothing runs until you approve it.

What you need

• A VPS or Linux box you can SSH into (Hetzner, DigitalOcean, AWS, a home lab—anything with bash)
• An Ohuriya account — sign up free and add credits when you are ready for AI tasks
• 30 seconds at a terminal on the server (plus a minute to sign up if you are new)

That is it. No API keys from OpenAI. No gateway wizard.

Step 1: Create your account

Open the Ohuriya app and sign up. You will land in the dashboard where servers and chat threads live. Credits deduct only when AI tasks run—not while your VPS sits idle. See how pricing works.

Step 2: Install the agent on your server

SSH into your VPS as a user with the permissions you want the agent to use (many people start with their deploy user, not root).

From the Ohuriya dashboard, copy the one-line install command for your server. It looks like:

curl -fsSL https://api.example.com/install.sh | bash -s -- ...

Paste it in your server terminal and run it. In roughly 30 seconds, the script downloads the agent (a compact binary), installs it, and connects it to your account. No manual config file editing.

Tip: If you use a firewall, allow outbound HTTPS from the server so the agent can reach Ohuriya.

Step 3: Confirm the server in the dashboard

Back in the web app, your server should appear as connected within seconds. If it does not show up, check:

• Install script finished without errors
• Server has outbound internet
• You used the install token from your account (not an old copied command)

Step 4: Your first chat—and your first approval

Open a chat for that server. Describe an outcome in plain English, not command syntax:

• “How much disk space is left?”
• “List running services”
• “Is nginx active?”

Ohuriya drafts a plan and shows exact commands before anything runs. Read them. If they look right, approve. If not, reject or ask for a revision.

That approval step is the product. ChatGPT would give you text to copy-paste. A raw autonomous agent might run first and explain later. Ohuriya waits for your OK.

Step 5: Try something slightly real

Once read-only checks feel comfortable, try a low-risk change you actually need:

• Restart a stuck service after you read the proposed command
• Clear old logs after you verify the path in the approval card
• Check SSL expiry dates before renewal season hits

Stay skeptical. Approval is not bureaucracy—it is how you keep production boring in a good way.

Troubleshooting quick hits

Agent not reachable — Re-run install from the dashboard token; confirm firewall outbound rules.

Permission denied on commands — The agent runs as the user you installed with; use sudo only when you intend to, and read those lines extra carefully.

Credits paused — Add credits in billing; your history and servers stay intact.

Why not stop at SSH?

You could. But when something breaks at night, you want context in one thread, proposed fixes you can reject, and no “I ran the wrong command on the wrong server” story. That is what Ohuriya is for.

Next steps

• Read why approve-before-run matters before you let anything touch production data
• Compare Ohuriya to OpenClaw and other agents if you are evaluating tools
• Browse the full feature comparison on the homepage

Connect once. Approve every step. Stop dreading the terminal.