Platform terms of service

These terms cover the Saanora Platform: the API and the console at this site, where you manage organizations, projects, API keys, usage and billing. By creating an account or sending a request, you agree to them. If you do not agree, please do not use the platform.

The privacy policy and the acceptable use policy are part of these terms. Saanora Chat, the consumer app at saanora.com, is a separate product with its own terms.

You must be an adult (18 or the age of majority where you live, whichever is higher) to use the platform.

If you use it for a company or another organization, you confirm that you have the authority to accept these terms on its behalf. In that case, "you" means both you and that organization.

You sign in with an email and password, or with Google. Keep your sign-in details safe and your account information accurate.

Work on the platform happens inside an organization. The person who signs up is responsible for their organization, including who they invite and which role each member gets: owner, admin, developer, billing or reader. Each organization has its own projects, keys, usage and credit.

You are responsible for everything that happens under your account and your organization. If you suspect someone has gained access who should not have it, tell us right away at support@saanora.com.

API keys start with sk_sa_ and belong to one project. Treat them like passwords:

• Keep them on your servers. Do not put them in browser code, mobile apps or public repositories.
• We store a SHA-256 hash of each key and its first few characters, so you can tell keys apart, but never the full key. That is why we cannot show a key to you again after it is created.
• Every request made with your key is charged to your organization, whoever sent it.
• If a key leaks, revoke it in the console and create a new one.

The platform is prepaid. You add credit in US dollars, and each request is paid from that credit.

• Top-ups start at $5. The largest single top-up depends on your usage tier: $25 on Tier 0 and Tier 1, $100 on Tier 2, $300 on Tier 3, $500 on Tier 4 and $1,000 on Tier 5.
• Polar is the merchant of record. It runs checkout, collects and remits taxes where they apply, and issues your receipts and invoices. Your card details go to Polar, not to us.
• Credit never expires. It is shared by every project in your organization and cannot be transferred to another organization.
• Before a request runs, its cost is held on your credit. When the answer comes back, the exact cost is deducted and the rest of the hold is released. If there is not enough credit, the request does not run. Failed requests are free.

Refunds go back to the original payment method through Polar. When a payment is refunded, the matching credit is removed from your balance and your usage tier is recalculated. To ask about a refund, write to support@saanora.com.

The details are in billing and credit.

Prices are per token and listed on the pricing page, which is always the current source. Today Clara costs $0.50 per 1M input tokens, and output is $0.

Fast mode, the default, bills the record and your questions read once. Consistent mode reads the record again with each question and bills it once for each question. A record longer than about 16K tokens is shortened in the middle and billed only for the part that was read. Every successful response reports the input tokens and the cost deducted for it. Failed requests are not charged.

We may change prices. A change applies only to requests made after it takes effect, never to requests already made. Prices for models that are not yet live, such as Mark, will be published before they open.

Rate limits apply to your organization as a whole. There are six usage tiers, set by your lifetime purchases and the days since your first payment. A request over your requests-per-minute or tokens-per-minute limit gets a 429 response with a Retry-After header telling you when to try again. See rate limits.

Each project can also have a monthly spend limit that you set. When a project reaches it, its requests are refused until the next calendar month (UTC) or until an owner or admin raises the limit. Because requests settle a few seconds after they run, spend can briefly go slightly over the limit.

Please do not try to get around limits, for example by spreading traffic across several organizations.

You must follow the acceptable use policy, which is part of these terms. It describes what the platform may not be used for.

You are responsible for how you use the platform and for how your own users use anything you build with it, and for following the laws that apply to you.

You own what you send: the record (the state), your questions and the options you define. You also own the answers the platform returns to you. We claim no rights in them.

You give us permission to process your content only to answer the request it came with. Records and questions are processed in memory to compute the answers and are not written to our database. Because we do not keep them, we do not use them to train models.

For each request we do store metadata so you can see your usage and so we can bill you: time, organization, project, key, endpoint, model, mode, status, number of questions, input and output tokens, latency, cost, request id and an internal billing reference. The full picture is in data handling and the privacy policy.

Make sure you have the right to send the data you send us.

Clara returns probabilities over the options you define. They are estimates, and they can be wrong. Before you let Clara make decisions on its own, validate it on your own labelled data and pick thresholds that fit your use. Its known limitations are listed in the model documentation.

You are responsible for the decisions you make with the answers, including whether a person should review them. The answers are not professional advice of any kind.

Our models are closed. Their weights are never released, and access through the API does not give you any rights to them.

You may not:

• Try to extract the model, its weights or its parameters.
• Distill it, which includes using its answers to train or improve another model.
• Reverse engineer, decompile or probe the models or the platform to work out how they are built.

The Saanora name, the Clara and Mark names, our design and our code are ours. Do not suggest an affiliation with us that does not exist.

We work to keep the platform fast and reliable, but we do not offer an uptime SLA and we do not promise it will never be interrupted.

The platform will change as we improve it. We may add models, change features, or deprecate and retire models and API versions. We announce changes and deprecations in the changelog.

You can stop using the platform at any time. To delete your account, write to support@saanora.com. There is no self-serve delete button.

We may suspend or end access for a key, a project, an organization or an account that breaks these terms or the acceptable use policy, puts the platform or other customers at risk, or where the law requires it. Where it is practical and lawful, we will warn you first.

The platform and the models are provided "as is" and "as available". To the maximum extent the law allows, we make no promises beyond what these terms state, including that the answers will be accurate or fit a particular purpose.

To the maximum extent the law allows, we are not liable for indirect or consequential damages, lost profits or lost data, or for decisions made based on the models' answers.

Our total liability for any claim is capped at the amount you paid for platform credit in the 12 months before the claim arose.

Some places do not allow some of these limits. Where that is the case, they apply as far as the law allows.

We may update these terms as the platform evolves. If we make a material change, we will update the date at the top of this page and note the change in the changelog. If you keep using the platform after that, you accept the updated terms.

Questions about these terms, a refund or your account: write to support@saanora.com.