Help centre

Everything from signing up to getting onto your machine. If something is still unclear, just ask us.

We sell two different things — check which one you need before ordering. Whole-machine rental gives you one exclusive machine you can install anything on and deploy your own models to. Rendering tasks put your job in a queue for our own pool of machines and hand you the result — you cannot host a service that way.

1. Whole-machine rental: from order to login

Create an account on the sign-in page, then top up with a voucher in your wallet. The full amount is held when you order, so the balance has to cover it.

Pick a machine and order

Choose a machine in whole-machine rental, pick hourly / daily / monthly, set the quantity and confirm. The machine is yours alone.

Download the helper

Machines have no public address — the connection runs over our tunnel. Download the relay for your system from the rental page or your rental details.

Configure ssh and you are in

Paste the ssh config snippet from your rental details into your local config file, then run ssh shulan.

2. How to reach it (the important section)

Machines come with no access address, no public IP and no web terminal. That is not a shortcut: the machine sits on its owner's home network and we have no public port to give it. The only way in is the SSH tunnel we hand you — your local ssh runs a small relay program that carries the byte stream through our platform to port 22 on your machine.

1. Download the relay

Pick your platform and put it somewhere you will remember (the examples below use ~/.shulan/):

Windows (64-bit) — file name shulan-ssh-windows-amd64.exe
macOS (Apple silicon)
macOS (Intel)
Linux (64-bit)

On macOS / Linux make it executable first: chmod +x shulan-ssh-darwin-arm64.

2. Paste this into your local `~/.ssh/config`

The rental id (--rental) and connection key (--key) are visible any time in the console under “My rentals” → Connect, so you can always copy them back; the key stays valid for the whole rental.

Host shulan
  HostName shulan
  User root
  ProxyCommand /Users/you/.shulan/shulan-ssh-darwin-arm64 --api https://www.shulan.net --rental 12 --key paste-your-connection-key-here
  ServerAliveInterval 30

On Windows point ProxyCommand at the exe instead, for example C:\Users\you\.shulan\shulan-ssh-windows-amd64.exe.

3. Connect

ssh shulan

You log in as root with the private key we generate — it is in your rental details too, and you can download it. You can also paste your own public key there instead, and then we stop storing a private key for you.

4. Want Jupyter in your local browser? Forward a port

ssh -L 8888:127.0.0.1:8888 shulan

Keep that ssh open and open http://127.0.0.1:8888 in your local browser — that is the service running on your machine. scp, sftp, rsync and VS Code Remote all work as well; they use the same channel.

3. What the page can do

Action	What it does	Watch out for
Power on	Brings the machine up, re-allocates the port, ssh works immediately	The owner's host has to be online; the button is greyed out when it is not
Power off	Shuts the machine down and stops the resource usage	Data stays; it is still there on the next boot
Restart	Equivalent to pulling the plug and putting it back	Anything running is interrupted
Reinstall OS	Rebuilds the machine from a system image — a clean machine	By default `/workspace` is kept; only ticking “also wipe /workspace” removes it

These buttons live on the instance page: in the console, click “Manage →” in that column of “My rentals”. A machine serves one tenant at a time, so the resources you see are all of them. Reinstalling does not move you to another machine, does not change your rental, and does not restart the billing.

4. Billing and refunds

Held on order: the amount held is unit price × quantity, moved from available to held balance the moment you order.
Early release: settled pro rata for the time used, and the rest goes back from held balance.
One unit minimum: anything under one unit is still charged as one unit — that is what stops “start it for a minute and kill it” games.
Provisioning failure on our side: refunded in full, nothing deducted. That one is our fault, not your bill.
Termination for violation: no refund. See section 6 below.

Every hold, settlement and refund leaves a line in your wallet ledger, precise to 0.000001.

5. How rendering tasks work

Rendering tasks run on our own pool of machines and are a different product from rentals: the machine is not yours, you submit a job, we queue and schedule it onto a machine, and you collect the result. That route cannot be used to deploy a model or host a service — for an exclusive machine, use whole-machine rental.

Pick a task type (video / image / audio / LLM / general) in the marketplace. The job then shows queued, running and finished, and the result is collected from the job page. Cancelling a job, or a queue timeout, refunds automatically.

6. Hard limits (we power off, we ban, we do not refund)

This section is not boilerplate. The platform detects these automatically. When it does, the machine is stopped at once, power-on is blocked, and we may terminate your rental with no refund at all; serious cases are banned and handed to the authorities.

No mining of any kind: no miner binaries (such as xmrig and friends), no mining pools, and no relaying mining traffic through your machine.
No unlawful content: nothing generated, stored or distributed that breaks the law or infringes anyone's rights.
No attacking: port scanning, password brute force, DDoS, breaking into other systems.
No using the machine as a hop: proxies, tunnel exits, anonymity nodes, or subletting it to someone else.
No dodging metering or billing: faking usage, or bypassing the platform's control channel.

The full text is in the terms of service. If you believe a violation was flagged wrongly, contact an administrator: after checking, they can clear the flag and let you power the machine back on.

7. Common questions

It will not connect — what to check first

Look at the machine's state under “My rentals”. A machine that is stopped / flagged / released is not supposed to connect.
Then check whether the owner's host is online. When it is off or offline the platform says so, and nobody can reach the machine until it is back.
Check that --rental and --key are copied correctly and belong to the same rental. If the key was regenerated on the page, your local copy is dead.
Still stuck? Run the relay on its own (not from inside ssh) — it prints the reason on screen: bad key, rental finished, machine not provisioned yet, all distinguishable.

Does reinstalling lose my data?

Anything in /workspace is kept by default; everything elsewhere is gone. Keep long-lived files in /workspace, not in /root or /tmp. Once the rental ends (released or expired) the machine and its data are destroyed together — we keep no backup.

Can I deploy my own models?

On a whole-machine rental, yes — the machine is exclusive, install whatever you need. On rendering-task machines, no: that route only runs submitted jobs.

Can I mine?

No. Mining in any form is detected automatically, the machine is stopped and the money is not refunded. There is no exception, and “just a little” is not a thing.

What if the owner's host goes offline?

Rental machines come from idle machines hosted by individuals, so a power cut, a dropped line or a shutdown on their side takes the machine away. The platform shows the host state honestly and greys the buttons out meanwhile. When the host is back the instance comes up by itself, with your data intact.

8. Still stuck?

Send us these and you will save most of the back-and-forth: the rental id, the exact error the relay printed, the full command you ran, and what you were running on the machine. Contact details are in the footer, or message an administrator.