Evaluation Agreement

1. Agreement, authority, and exact version

This Evaluation Agreement is between AI Gateway HQ LLC, doing business as AI Gateway HQ, and the person or organization named in an AI Gateway HQ pilot invitation (“Participant”). It applies only to the invitation that identifies version evaluation-2026-08-10-v1. If you accept for an organization, you represent that you are authorized to bind that organization.

Acceptance is recorded against the signed-in identity, organization, invitation, and immutable agreement version. An invitation is personal to its named recipient and may not be transferred, shared, sold, or used to create access for an unapproved person.

2. Evaluation purpose and scope

AI Gateway HQ grants Participant a limited, nonexclusive, revocable right to evaluate the enabled Services for internal testing and feedback during the invitation period. The invitation states the expiration, participant allocation, enabled features, permitted model targets, request and token limits, concurrency ceiling, and any sponsored-credit ceiling.

Evaluation access is not a production commitment, a generally available service, a service-level agreement, a promise that a feature will remain available, or authorization to use an unlisted deployment, model, data class, integration, or regulated workflow.

3. Costs, provider accounts, and hard limits

• Bring-your-own-key provider inference remains governed and billed under Participant's separate provider relationship.
• Sponsored or managed model use is limited to the exact feature, model, expiration, and monetary ceiling shown in the invitation and available prepaid entitlement.
• Promotional or sponsored credit has no cash value, cannot be transferred, and ends when exhausted, expired, withdrawn, or revoked.
• The Services may reserve usage before forwarding a request and must stop eligible billable traffic when the applicable hard boundary or usable balance is exhausted.
• No card is charged merely by accepting an evaluation. A paid purchase or auto-reload requires a separate affirmative checkout or configuration action showing the commercial terms.

4. No SLA and evaluation changes

The evaluation is provided “as is” and “as available,” without a production SLA, uptime commitment, response-time commitment, support-response guarantee, feature-continuity promise, or disaster-recovery commitment. Planned maintenance, provider incidents, security response, abuse prevention, cost controls, and product changes may interrupt access.

AI Gateway HQ may add, modify, rate-limit, suspend, or remove evaluation functionality. When practical, material changes will be communicated through the account or invitation contact. Participant should not make the evaluation a dependency for production, safety-critical, or time-critical operations.

5. Participant responsibilities and prohibited use

• Protect sign-in factors, gateway keys, endpoints, and provider credentials; use a separate gateway key for each workload; and revoke access promptly after suspected compromise.
• Use only provider accounts, content, and data that Participant is authorized to use, and comply with provider terms and applicable law.
• Do not attempt to evade limits, probe another tenant, disrupt the Services, extract nonpublic software or controls, introduce malicious code, or use the evaluation to build or benchmark a competing service from nonpublic elements.
• Do not submit protected health information, payment-card data, government-classified or export-controlled information, or other specially regulated or highly sensitive data unless a written Order expressly permits the exact use and required controls are active.
• Do not use model output as the sole basis for a consequential decision or as a substitute for qualified professional judgment. Participant is responsible for appropriate human review and validation.

6. Security, content, and third-party models

Participant controls its users, workloads, provider choices, keys, routes, policies, budgets, and data submitted during the evaluation. AI Gateway HQ applies only the controls described as implemented for the enabled evaluation. No control eliminates all security, model, provider, endpoint, or human risk.

Model requests are sent to the configured provider. Third-party model behavior, output quality, availability, processing location, data handling, and independent charges remain subject to that provider's terms. Participant must retain any records it is required to keep and must not rely on the evaluation as its sole backup or system of record.

7. Confidentiality, feedback, and ownership

Each party will protect the other party's nonpublic information with reasonable care and use it only for the evaluation. This does not cover information independently developed, lawfully received without restriction, publicly available without breach, or approved for release in writing.

Participant retains its rights in Customer Content. AI Gateway HQ retains its rights in the Services, documentation, designs, and nonpublic technology. Voluntary feedback may be used to improve the Services without identifying Participant or disclosing Participant Confidential Information.

8. Suspension, revocation, and expiration

AI Gateway HQ may immediately suspend or revoke affected access to enforce an invitation boundary, contain a security incident, prevent harm or fraud, address provider or payment risk, comply with law, or investigate a material violation. Participant may stop using the evaluation at any time.

Access ends automatically at the invitation expiration, when its limits are exhausted where applicable, or when revoked. Expiration or revocation does not create a right to replacement credit, continued access, source code, production support, or a paid contract. Payment, confidentiality, ownership, disclaimers, and liability provisions survive where their nature requires.

9. Disclaimers and liability boundary

To the maximum extent permitted by law, AI Gateway HQ disclaims implied warranties of merchantability, fitness for a particular purpose, title, and noninfringement. AI output can be incomplete, inaccurate, or unsafe. AI Gateway HQ does not guarantee a specific result, savings amount, certification outcome, provider availability, or compatibility that is not expressly listed as validated.

The exclusions and liability boundaries in the Terms of Service apply. For a free evaluation, AI Gateway HQ's aggregate liability is capped at one hundred U.S. dollars, except where a limitation is prohibited by law. Neither party is liable for indirect, incidental, special, consequential, exemplary, or punitive damages, or lost profits, revenue, goodwill, or data, to the maximum extent permitted by law.

10. Incorporated policies and later versions

The Terms of Service, Acceptable Use Policy, Privacy Notice, and Refund and Cancellation Policy are incorporated into this Evaluation Agreement. If this Evaluation Agreement expressly conflicts with those documents for the bounded evaluation, this Evaluation Agreement controls for that conflict. An executed Order controls where it expressly says otherwise.

A future material revision will use a new immutable agreement identifier. Updating the public text does not silently alter the version already recorded for an accepted invitation. Participant must accept the version presented with any new invitation before new evaluation access is granted.