Terms of Service

1.Who we are and how you accept

“ScaledThought”, “we” and “us” mean ScaledThought Inc., of Vancouver, British Columbia, Canada. “You” and “the customer” mean the person or organisation on whose behalf an account is created.

An account can be created for you by you or by an AI agent, script or other software (an “agent”) through the CLI (init) or the API (POST /v1/accounts). Creating it accepts nothing on your behalf. You accept these terms and the Credits & Refunds policy yourself, on the payment page, when you first buy credits or save a card for auto-reload; we record the version you accepted then, and again with each later payment. Until then the account can't be used for anything we charge for.

If you accept on behalf of an organisation, you confirm you are authorised to bind it. You must be of the age of majority where you live and able to enter into a binding contract.

2.Agents and the humans they act for

• The customer is always the human or organisation the agent acts for, never the agent. An agent can create an account and keys for you, but it can't accept these terms for you: you do that on the payment page (our setup instructions at /llms.txt tell agents so).
• You are responsible for everything done with your account and your API keys, including by your agents and by anyone who obtains your keys, until you revoke them.
• Payments need a human. An agent can only request a top-up; it receives a payment link, and nothing is charged until a human opens that link, accepts these terms and the Credits & Refunds policy, and pays. Approving a payment link, including any auto-reload terms on it, is your authorisation of that payment.
• If you give an agent a key with the billing scope, it can create payment links and lower or turn off auto-reload, but it can't enable auto-reload or raise its limits without a new human approval.

3.The service

ScaledThought provides headless, OpenAI-compatible inference on open-weight models: chat and text completions, embeddings, an automatic model router, batch inference, LoRA fine-tuning, API keys with budgets and rate limits, usage and billing information, an MCP server, a CLI and a dashboard where the people on your account sign in by email to see billing and keys, manage members and connect agents.

Home region. Every account belongs to one home region, chosen when it is created and fixed for its lifetime: us (the United States, AWS us-east-2) or ca (Canada, AWS ca-central-1). Your API keys carry the region (st_us_… or st_ca_…). We process and store your content and account data in that region, as described in the Privacy Policy. Card payments are processed by Stripe in the United States for both regions.

Models change. The catalog on our website is not a promise that a model is available. Which models are live in a region can change at any time; list them with GET /v1/models. We may add, change, rate-limit or retire models, features and API versions. We will try to give reasonable notice of changes that break working integrations.

Availability. We work to keep the service fast and available, but it is provided without a service-level agreement. Features described as preview or beta, and fine-tuning, may change or be withdrawn.

4.Accounts, API keys and security

• Keep your API keys secret. Don't put them in client-side code, commit them to source control or paste them into chats. The CLI stores them in a local .env file with restricted permissions and adds it to .gitignore.
• We store API keys only as keyed hashes (HMAC-SHA-256), so we can't show you a key again or recover it. If a key is lost or exposed, revoke it (DELETE /v1/keys/{id}) and create a new one.
• Accounts that haven't bought credits can have up to 5 active keys. Don't create multiple accounts to get around limits; see the Acceptable Use Policy.
• Tell us promptly at security@scaledthought.com if you believe your account or keys have been compromised.

5.Credits and payment

Usage is paid from prepaid credits, bought in US dollars through a payment link a human approves ($20 to $2,000 per purchase, at most $2,000 in any 24 hours), optionally with auto-reload on amounts the human approves. When credits run out, requests are refused with HTTP 402; we don't send bills after the fact.

Usage is charged at the per-model prices published on our pricing page at the time of the request, measured from the token counts reported by the model server. Price changes apply to requests made after they are published; they don't change the dollar value of credits you already hold.

The Credits & Refunds policy is part of these terms and sets out how credits, auto-reload, refunds, taxes and disputes work.

6.Your content

“Your content” means the prompts, files, training data and other inputs you send, and the completions, embeddings, batch results and fine-tuned models the service produces for you.

• As between you and us, you own your content. We claim no rights in it beyond what we need to run the service for you.
• You grant us a limited licence to process your content only to provide the service to you: to run your requests, batches and fine-tuning jobs, to store what you ask us to store, and to meet legal obligations.
• We don't use your content to train, fine-tune or evaluate any model, except to run fine-tuning jobs you request, whose results are private to your account.
• We don't log or store the prompts and completions of ordinary API requests. Batch and fine-tuning files are stored until you delete them or for 30 days, whichever comes first. The Privacy Policy has the details.
• You are responsible for your content: you must have the rights and permissions needed to send it to us (including any notices to, and consents from, the people it is about) and for it to be processed in your home region as described in the Privacy Policy.

7.AI output

Models can produce output that is wrong, incomplete, offensive, biased, or similar to existing material. Output is not reviewed by us. You are responsible for evaluating it and for how you and your agents use it, including checking it before relying on it for decisions that affect people's rights, health, safety, finances or legal position.

Benchmark scores and router choices are estimates, not guarantees of quality for your task.

8.Model licences

The models we serve are open-weight models published by third parties under their own licences, some with their own acceptable-use terms, attribution requirements or limits. You must comply with the licence of each model you use, as linked from its model page and summarised in the Acceptable Use Policy. The model publishers are not parties to these terms and give no promises through us.

9.Acceptable use

You must follow the Acceptable Use Policy and make sure your agents and your own users do too. You must not use the service to break the law, to harm others, to attack or probe our systems (other than under our vulnerability disclosure policy), or to get around its limits.

10.Suspension and termination

• You can stop using the service at any time, revoke your keys, and ask us to close your account at support@scaledthought.com.
• We may suspend an account or key immediately if we reasonably believe it breaks these terms or the Acceptable Use Policy, is compromised, is the subject of a payment dispute, puts the service or other customers at risk, or if the law requires it. Suspended accounts' requests are refused.
• We may end these terms for any account on 30 days' notice. If we do so for reasons other than your breach, we refund your unused purchased credits.
• When an account is closed, we delete or retain its data as described in the Privacy Policy. Refunds of unused credits on closure follow the Credits & Refunds policy.

11.Disclaimers

To the extent the law allows, the service, the models and their output are provided “as is” and “as available”, without warranties of any kind, express or implied, including merchantability, fitness for a particular purpose, accuracy and non-infringement. Nothing in these terms limits rights you have under consumer protection laws that can't be excluded.

12.Limitation of liability

To the extent the law allows:

• neither party is liable for indirect, incidental, special, consequential or punitive damages, or for lost profits, revenue, data or goodwill, arising out of these terms or the service;
• each party's total liability arising out of these terms or the service is limited to the greater of the amounts you paid us in the 12 months before the event giving rise to the claim and US$100.

These limits don't apply to your payment obligations, to your indemnity below, or to liability that can't be limited by law.

13.Indemnity

You will defend and indemnify us against third-party claims arising from your content, your use of output, or your (or your agents') breach of these terms, the Acceptable Use Policy or a model licence, and pay resulting damages and reasonable costs.

14.Governing law and disputes

These terms are governed by the laws of the Province of British Columbia and the federal laws of Canada that apply there, without regard to conflict-of-laws rules. The courts of Vancouver, British Columbia have exclusive jurisdiction over disputes arising out of these terms, and both parties submit to them.

If you are a consumer, you may also have the right to bring proceedings in the courts where you live, and the mandatory consumer protection laws of the place where you live continue to apply.

15.Changes to these terms

We may update these terms. Each version has a date-based version string, shown at the top of this page, and the API and CLI return the current version when an account is created. We will post material changes here and in /llms.txt at least 30 days before they take effect, and by email if you have given us one. Continuing to use the service after a change takes effect means you accept it; if you don't, stop using the service and ask for a refund of unused purchased credits under the Credits & Refunds policy.

16.General

• These terms, with the policies they incorporate, are the whole agreement between us about the service and replace earlier agreements on the same subject.
• You may not assign these terms without our consent. We may assign them to an affiliate or to a successor of our business, with notice to you.
• If part of these terms is unenforceable, the rest stays in effect. Not enforcing a right is not a waiver of it.
• Neither party is liable for delays caused by events beyond its reasonable control.
• Legal notices to us go to legal@scaledthought.com; ask there if you need our postal address. We send notices to the email on your account, if any, and otherwise post them on this site and in /llms.txt.

17.Contact

Questions about these terms: legal@scaledthought.com. Support: support@scaledthought.com. Security reports: security@scaledthought.com. Privacy: privacy@scaledthought.com.