Terms of use

Noctrel is built and operated by an individual developer, not a company. “We” below means that person. There is no legal entity, no registered office, and no group of companies behind it.

Using this website means you accept these terms. Creating an account and using the Noctrel console means the same.

Noctrel is early-stage software, built and run by one person. That is not a marketing word here — it means the software is incomplete, actively changing, and can break. Specifically:

• There is no uptime commitment and no service level agreement.
• Features can change or be removed without notice.
• Access can be granted, paused or withdrawn at any time.
• Data can be lost. Keep your own backups of anything you cannot lose.

The service is provided “as is”, without warranties of any kind.

Every reply in the console’s chat, every self-check, and every proposed command comes from an AI model — never a person on our side reading and responding. That is stated here explicitly, not just left implicit in the product, in line with the EU AI Act’s transparency requirement (Article 50), in force since 2 August 2026.

Treat what it says the way you would a junior engineer’s suggestion: worth reading, not worth trusting blindly. That is exactly why the approval gate described below exists.

This is the part worth reading twice. You install an agent on machines you control. That agent runs commands on those machines, sometimes as root. An AI model decides which command to propose.

The design is built so that you stay in control of that, and you should understand how:

• The AI can only propose commands — either from the catalogue you wrote, or, when nothing in it fits, a single new one-off command it drafts for that job. Either way, it never gets a free-form shell: nothing it proposes runs until a human approves it.
• Anything flagged as risky, or requiring root, waits for a human to approve it — unless you raise that server’s autonomy level, which is a setting only you can change.
• Every proposal, approval, denial and result is recorded in an audit log you can export.

Within that, you remain responsible for what runs on your infrastructure. You write the catalogue, you choose the autonomy level, and you press Approve. A model can be wrong, and a command you authorised can still break something. Do not point this at production you cannot afford to have interrupted until you trust it on something you can.

Keep your credentials to yourself, and turn on two-factor authentication. You are responsible for what happens under your account, including anything a teammate you invited does with the access you gave them.

Tell us promptly if you think an account or an API token has been compromised.

By default, chat runs on a Groq key Noctrel itself provides — you do not need to bring your own to get started. You pick which curated model your org uses under Settings, and that usage counts against the limits on your plan.

If you would rather use your own provider and model (Anthropic, OpenAI, Google, or your own Groq key), you can add that under Settings — once set, it always takes over from Noctrel’s default. Calls made with your own key are billed to you by that provider, at their rates — including calls the AI makes on its own schedule if you enable that. Noctrel shows you your usage against the limits on your plan, but the provider’s bill is between you and them.

• Use it against machines you are not authorised to administer.
• Use it to break the law, or to attack, disrupt or gain access to systems that are not yours.
• Try to defeat the approval gate, the risk classifier, or the audit log.
• Resell access, or share your account with people outside your organisation.

Access can be withdrawn immediately if any of this happens.

The Free plan is free — no payment method is collected and no card is charged. Pro is coming and not yet chargeable; any paid-looking plan shown in the console today is simulated. When Pro actually launches, existing users will be told before anything becomes chargeable, and nobody will be billed retroactively for time already spent on Noctrel.

The software, the interface and the brand stay ours. Your data — your servers, your catalogue, your conversations, your audit log — stays yours, and you can export all of it from the console at any time.

To the fullest extent the law allows, we are not liable for lost data, lost revenue, downtime, or consequential damage arising from use of this free, early-stage service. Nothing here tries to exclude liability that cannot legally be excluded — including for death, personal injury, fraud, or anything else the law reserves.

You can stop at any time and request deletion from the console. That starts a 90-day grace period, not an instant wipe: your account keeps working exactly as before, you get an email explaining the date it becomes permanent, and you can cancel any time before then from Settings › Security. After 90 days it is permanently and irreversibly deleted.

We can end access at any time, particularly if Noctrel shuts down or these terms are broken. If we end it without cause, you will be given a reasonable chance to export your data first. Access ended by us for cause (fraud, abuse, a banned account) is not subject to the 90-day grace period above — that window exists for your own decision to leave, not to keep a removed account’s data around by default.

Noctrel is operated from Türkiye, and these terms are governed by Turkish law. Turkish courts have jurisdiction over any dispute arising from them.

If you are a consumer somewhere else, this does not take away rights your own country’s law gives you and does not let us contract out of them — it only settles which law applies where there is a choice.

These terms will change as the product does. The date at the top changes with them, and anyone with an account is emailed about changes that materially affect them.