TitanEnsemble

TitanEnsemble — Terms of Service

Last updated: 17 August 2026

1. Who these terms are between

These terms are an agreement between you ("you", "your") and Realm Intelligence, 2711 Centerville Road, Suite 400, Wilmington, Delaware 19808, United States ("TitanEnsemble", "we", "us"). By creating an account or using the service you accept them.

TitanEnsemble is a service of Realm Intelligence. Realm Intelligence operates the platform under an exclusive license from Liminal Intelligence LLC, which owns the underlying intellectual property. We may assign these terms to a wholly-owned subsidiary operating the TitanEnsemble service, on notice to you and with no change to your rights.

2. What the service is

TitanEnsemble is an intelligence layer over language models. You subscribe to the layer — the memory, knowledge, threading, routing and tooling around a model — not to the tokens.

Two ways a request is served:

  • Bring your own key (BYOK). You store a credential for a supported provider, and requests for the models on that provider run on your credential. You are billed by that provider, at their price. We do not mark up, resell, or intermediate that billing relationship.
  • Hosted seats. Some models run on our own compute. These are included in your subscription and reach no third-party provider.

Supported providers are a curated set, currently OpenRouter, Together AI and Novita. We may add or remove providers; removal does not delete your stored credential, but requests to that provider will stop.

3. Accounts and API keys

  • You are responsible for activity under your account and keys, and for keeping keys secret.
  • An API key may carry its own conversation and its own configuration. Deleting a key ends access through it.
  • You must be able to form a contract in your jurisdiction, and you must not share credentials with anyone you would not give your account to.

4. Your provider credentials

This section is the heart of a BYOK product, so it states what is technically true and no more.

What we do with a credential. Exactly one thing: dispatch your requests to the provider you chose. This is structural, not a policy — the BYOK dispatch path carries no credential of ours, so your traffic cannot silently fall back onto a house account.

How it is held. Encrypted at rest in a secret vault, addressed by a derived per-tenant path, reached with tenant-scoped broker tokens. Cross-tenant reads are refused by the store's own policy, not merely by our application code.

No read-back. No API of ours returns a stored credential value — including to us. Retrieving one would require deliberate engineering work, and such work is auditable.

⚠️ What we deliberately do NOT claim. We do not claim that no person inside the company could ever reach a credential. No system with an operable secret store can honestly say that: the store has a root credential used for bootstrap, and a person with production access and that credential could read tenant paths. We would rather tell you the true shape of the control than publish a stronger sentence we could be held to. (Nor do we claim a value is "never readable back" in an absolute sense — the service must be able to decrypt it to make your calls.)

Deletion. You may delete a stored credential at any time; it is removed by stored reference, and again during account teardown. If any part of a deletion fails, the report says so rather than reporting success.

5. Your content

You own your content — prompts, attachments, threads, knowledge sets, briefings and the memories derived from them. We claim no ownership of it and no licence to it beyond what is needed to run the service for you.

Where it goes is documented in the privacy page and summarised here: your prompts and context go to the provider you selected and nowhere else; hosted-seat requests stay on our own compute. Memories, knowledge sets, briefings, threads and usage records do not leave the platform.

We do not train on your content.

You are responsible for what you send — including that you have the rights to it, and that sending it to your chosen provider is lawful for you. Your use of a provider is also subject to that provider's terms.

6. Acceptable use

You may not use the service to:

  • break the law, or infringe someone's rights;
  • generate sexual content involving minors, or content whose purpose is to harass, defame or threaten a person;
  • build weapons, malware, or tooling whose primary purpose is unauthorised access to systems;
  • attempt to breach our isolation boundaries — other tenants' data, the secret store, or the control plane;
  • resell raw model access in a way that makes us an undisclosed intermediary between the provider and an end user, contrary to that provider's terms.

The models you reach also carry their providers' policies, and those apply to you as well.

7. Subscription, payment and usage

  • The subscription covers the intelligence layer. Provider inference on your own key is billed by that provider, to you, directly.
  • Hosted-seat usage is included, subject to the quotas published on the pricing page.
  • Fees are charged in advance for the period and are non-refundable except where required by law, or where we have failed to provide the service and you tell us within a reasonable time.
  • Taxes are calculated at checkout where required.
  • We may change prices. Changes take effect at your next renewal, and we will tell you before they do.

8. Availability, and what we do not promise

We aim for the service to be available and correct, and we do not promise that it always will be.

Specifically we do not warrant that: model output is accurate, suitable, or fit for any purpose; a third-party provider will remain available or keep its prices; or that the service will be uninterrupted. Model output is generated text — you are responsible for reviewing it before you rely on it, and especially before relying on it in a regulated, medical, legal or financial setting.

The service is provided "as is" to the maximum extent the law allows.

9. Limitation of liability

To the maximum extent permitted by law, neither party is liable for indirect, incidental, special or consequential damages, or for lost profits or lost data, arising from these terms.

Our total liability for any claim is limited to the amount you paid us in the twelve months before the event giving rise to it.

⚠️ This limit is measured against our fees. It does not extend to what a provider bills you on your own credential, because we are not a party to that transaction — which is the same reason we do not mark it up.

Nothing here excludes liability that cannot lawfully be excluded.

10. Indemnity

You will defend and indemnify us against third-party claims arising from your content, your use of the service in breach of these terms, or your breach of a provider's terms.

11. Suspension and termination

You may stop at any time by deleting your account; deletion runs the teardown described in §4 and the privacy page.

We may suspend or terminate an account for a serious or repeated breach of §6, for non-payment, or where required by law. Where circumstances allow, we will tell you first and give you a chance to fix it.

On termination: your subscription ends, your data is deleted per the retention rules in the privacy page, and audit records are retained to the compliance floor that applies to your tenant's regime — a deletion request cannot shorten a regulatory retention floor, and we say so rather than implying otherwise.

12. Changes to these terms

We may update these terms. Material changes will be notified before they take effect, and the "last updated" date above will change. Continuing to use the service after that is acceptance.

13. Governing law and disputes

These terms are governed by the laws of the State of California, United States, without regard to its conflict-of-laws rules. The courts located in San Francisco County, California have exclusive jurisdiction over any dispute arising from them, and both parties consent to venue there.