Terms of service

The operator details this page depends on (publisher identity, hosting provider and contact addresses) have not been published yet. This text is a draft for review. It is not indexed by search engines and does not take effect until those details are published.

These terms govern your use of gpuos.si, the gpuOS dashboard and the hosted gpuOS gateway (the service). By creating an account, creating an API key or using the service, you agree to them. If you do not agree, do not use the service.

The service is meant for professional use by developers and teams. By registering you confirm that you act for a business or in a professional capacity and, if you act for an organisation, that you may bind it. The legal notice identifies the publisher, and the privacy policy explains how personal data is handled.

gpuOS connects GPU machines that you operate to a hosted, OpenAI-compatible API. You install the gpuOS agent on your machines, deploy catalog models with Ollama and call the hosted endpoint with workspace API keys. The gateway checks the key and plan limits, relays the request to one of your online nodes and returns the response. Inference runs on your hardware.

The service does not include:

• renting GPUs for you;
• routing requests to external model providers;
• offline or air-gapped operation, because the gateway needs internet connectivity;
• an availability commitment or a support response time.

Pro is a subscription paid through Stripe. The quantity follows the number of GPUs your nodes report, with a minimum of one. It renews each billing period until you cancel it. Prices are in US dollars and exclude VAT and other taxes, which are added where they apply.

You can cancel from the billing portal in your workspace settings. Cancellation takes effect at the end of the period you have paid for. Fees already paid are not refunded, except where the law requires it or where we agree otherwise in writing.

If a subscription ends or is no longer paid, the workspace returns to the Community plan and its limits. We may change prices or plan limits for future billing periods. A change does not affect a period you have already paid for.

• Give accurate information and keep your sign-in credentials safe.
• gpuOS and cpuOS share one account, with the same workspaces, members and roles on both, and the member limit of the workspace's gpuOS plan (2 on Community, pending invitations included) applies on both. API keys are per product by default: a gpuOS key does not call cpuOS unless you allow it to call both products when you create it. Revoking a key that works with both products stops it on both; the gpuOS gateway can keep accepting it for up to 5 seconds after it was revoked on cpuOS. Using cpuOS is governed by its own terms.
• Workspace owners and admins decide who joins and which keys exist. You are responsible for what happens under your workspace's keys and node tokens.
• API keys and node tokens are shown once. Treat them as secrets, revoke or rotate them if they leak, and keep them out of client-side code and public repositories.
• Tell us promptly at To be provided if you think an account was compromised.

You provide, secure and pay for the machines you connect, and you must be authorised to use them. The agent runs with the permissions you give it when you install it. Hardware, hosting, power and network costs are yours.

The catalog lists open models published by third parties under their own licences. We grant no licence to those models. Check the licence on each model page and make sure your use complies with it. The memory figures shown before deployment are estimates. They do not promise that a model will fit or perform on your hardware.

You must not use the service to:

• create, process or distribute content that is illegal where you or your users are located, or that infringes someone else's rights;
• attack, scan or probe the service, other users' workspaces or third-party systems, or try to reach data or accounts that are not yours;
• connect a machine you do not control or are not authorised to use as a node;
• bypass or tamper with rate limits, quotas, plan limits or usage metering, or create accounts to evade them;
• send traffic designed to degrade the service for others;
• create malware or run phishing, fraud or spam campaigns, or harass people.

You are responsible for the prompts you send, the outputs you use and your own users' use of your application.

We may rate-limit, suspend or remove access, or delete a workspace, if we reasonably believe these rules are broken or that the service, other users or third parties are at risk. Where practical, we tell you why. To report abuse, write to To be provided.

You keep your rights in your prompts, outputs, code and data. You give us only the permission needed to operate the service for you: to relay your requests and responses, to meter usage and to store your account and workspace data.

We do not store prompt or completion text, so we do not use it to train models. The privacy policy lists what is stored.

Models can produce wrong, incomplete or inappropriate output. You choose, configure and run the models on your own hardware. Review output before relying on it, especially for decisions that affect people. You are responsible for how you use it.

gpuOS is an early product. It comes without an availability commitment, a service level agreement or a guaranteed support response time. The hosted gateway can restart or be unavailable, which interrupts requests in progress. Monthly token limits are checked when a request is admitted, so concurrent requests can pass the limit.

We may change, add or remove features, catalog models, plan limits and agent versions.

You can stop at any time: cancel any subscription, remove your nodes, revoke your keys and delete your account in Settings > Your account. The privacy policy says what is deleted. The account is shared with cpuOS, so deleting it removes it there too. Workspaces shared with other people stay, and their only owner must hand over ownership first. For anything else, use the data request contact in the privacy policy.

We may suspend or end your access if you break these terms, if your use puts the service or others at risk, or if the law requires it. When access ends, your keys stop working and your nodes stop receiving requests.

To the extent the law allows, the service is provided as is and as available, without warranties of any kind, including fitness for a particular purpose and uninterrupted or error-free operation.

Nothing in these terms limits liability that the law does not allow us to limit, including liability for fraud, intentional misconduct, gross negligence, and death or personal injury.

Subject to that, we are not liable for indirect or consequential loss, or for loss of profit, revenue, data or goodwill. We are not liable for loss caused by your machines, models or applications, or by third-party services such as Stripe, GitHub, Google or model licensors. Our total liability for all claims relating to the service is limited to the amount you paid us for the service in the 12 months before the event that gave rise to the claim.

We may update these terms. The date at the top of this page shows the latest revision (2026-10-15). Changes apply from the date they are published. If you keep using the service afterwards, you accept them. If you disagree, stop using the service and cancel any subscription.

These terms are governed by To be provided. Disputes go to To be provided, to the extent mandatory law allows.

• Legal notice
• Privacy policy