legal

Terms of Service

Last updated: March 25, 2026

1. Service Description

FGY Cache ("the Service") is a managed proxy layer that intercepts LLM API traffic and serves cached responses to reduce redundant inference costs. The Service is operated by FGY ("we", "us", "our") and is available at fgy.ai and related subdomains.

The Service provides exact hash matching, pgvector-based semantic similarity matching, BEAM request coalescing, and full streaming support across supported providers including OpenAI, Anthropic, and OpenRouter.

2. Account and Access

To use the Service you must create an account via our dashboard. Authentication is handled through Clerk, a third-party identity provider. You are responsible for maintaining the security of your account credentials and API keys.

Each account provisions a tenant on the FGY cluster. You may generate multiple API keys scoped to your tenant, each with optional monthly usage limits and expiry periods. API keys are shown once at creation and cannot be retrieved afterward.

3. Billing and Payment

FGY operates on a prepaid credit model. You load credits to your balance via Stripe-powered checkout sessions. The Service charges a percentage of the provider cost avoided on each cache hit. Cache misses are never billed.

Charges are calculated using per-model, per-provider token rates and deducted from your prepaid balance in real time. If your balance reaches zero, cache hits that would incur a charge will pass through to the upstream provider at your own cost instead.

You may configure automatic top-ups via Stripe to maintain a minimum balance. Credit codes may be redeemed for one-time balance additions. All balance mutations are recorded in an append-only ledger visible in your dashboard.

4. Provider API Keys

FGY does not store, manage, or persist your LLM provider API keys. Your provider bearer token is sent with every request via the X-Provider-Auth header and is forwarded upstream only on cache misses. On cache hits, your provider key is never transmitted.

You are solely responsible for your provider API keys, their associated costs, and compliance with your provider's terms of service.

5. Cached Content

When requests pass through FGY, prompt-response pairs are cached within your tenant's isolated namespace. Cached entries include prompt hashes, response bodies, token counts, and pgvector embeddings for semantic matching. Cache entries are scoped to your tenant and are not shared with other users.

You retain ownership of all content that passes through the Service. We do not use your cached prompts or responses to train models, build products, or for any purpose beyond operating the caching service for your tenant.

6. Availability and Limitations

The Service is provided "as is" without warranty of any kind. We make reasonable efforts to maintain uptime and data integrity but do not guarantee uninterrupted availability. The Service acts as a proxy layer — upstream provider outages, rate limits, and errors are outside our control and are passed through as-is.

We reserve the right to modify, suspend, or discontinue the Service with reasonable notice. Material changes to pricing will be communicated at least 30 days in advance.

7. Acceptable Use

You agree not to use the Service to cache or proxy content that is illegal, harmful, or violates the terms of your upstream LLM provider. You agree not to attempt to access other tenants' cached data, reverse-engineer the caching infrastructure, or deliberately circumvent billing mechanisms.

8. Termination

Either party may terminate this agreement at any time. Upon termination, your cached data will be purged within 30 days. Any remaining prepaid credit balance at termination is non-refundable unless otherwise required by applicable law.

9. Limitation of Liability

To the maximum extent permitted by law, FGY shall not be liable for any indirect, incidental, consequential, or punitive damages arising from your use of the Service, including but not limited to lost profits, data loss, or costs incurred from upstream provider charges on cache misses.

10. Changes to Terms

We may update these terms from time to time. Changes will be posted on this page with an updated effective date. Continued use of the Service after changes take effect constitutes acceptance of the revised terms.

Questions about these terms? Reach us at [email protected].