← Back to ProxyLLM

Terms of Service

Last updated: June 18, 2026

These Terms of Service ("Terms") govern your access to and use of ProxyLLM (the "Service"), an LLM proxy platform that sits between your applications and third-party AI providers, offering semantic caching, smart model routing, cost tracking, and a management dashboard. The Service is operated by Sysdev TechStrategy & Consulting ("ProxyLLM", "we", "us", or "our"), a company organized under the laws of Brazil. Our full legal identity, registration number (CNPJ), and registered address are set out in Section 18 (Contact).

By accessing or using the Service, by clicking to accept these Terms, or by creating an account, you ("you", "Customer", or "User") agree to be bound by these Terms. If you are entering into these Terms on behalf of a company or other legal entity, you represent that you have authority to bind that entity, and "you" refers to that entity. If you do not agree to these Terms, you must not access or use the Service.

These Terms incorporate by reference our Privacy Policy and our Subprocessors list, and, where applicable, a Data Processing Addendum (available on request at contact@proxyllm.dev).

1. Definitions

  • "Service" means the ProxyLLM proxy API, dashboard, landing site, documentation, and all related software and infrastructure operated by us.
  • "Upstream Provider" means any third-party AI model or embedding provider to which the Service forwards requests, including OpenAI, OpenRouter, and Anthropic.
  • "Inputs" means the prompts, system messages, parameters, files, and other content you transmit through the Service to an Upstream Provider.
  • "Outputs" means the completions, responses, embeddings, and other content generated by an Upstream Provider and returned through the Service.
  • "Customer Content" means, collectively, your Inputs and Outputs.
  • "BYOK" ("Bring Your Own Key") means the optional feature, available on paid plans, that lets you configure your own Upstream Provider API keys so that requests are sent directly under your own provider account.
  • "Managed Mode" means the default mode in which requests are routed through our own Upstream Provider account (via OpenRouter), subject to a per-tier model whitelist.
  • "Workspace" means the logical tenant boundary to which your account, API keys, logs, cache, and configuration are scoped.
  • "Subprocessor" means a third party engaged by us to process personal data or Customer Content in connection with providing the Service.

2. Description of the Service

ProxyLLM is an API proxy that forwards your requests to Upstream Providers and returns their responses. We provide an OpenAI-compatible endpoint (/v1/chat/completions), an Anthropic-compatible endpoint (/v1/messages), semantic and exact-match caching, configurable model routing, usage and cost analytics, and a dashboard.

We do not provide, train, own, or control the underlying AI models. The models are operated by independent Upstream Providers under their own terms. We act solely as a passthrough and optimization layer. Model availability, behavior, pricing, content policies, and output quality are determined by the Upstream Providers, not by us.

In Managed Mode, the available models are limited to a per-tier whitelist that we may change at any time. As of the last-updated date, the Managed Mode whitelist on every tier (Free, Pro, and Scale) is limited to a single model, gpt-4o-mini. Anthropic Claude models (and any other model not on the current Managed Mode whitelist) are not available in Managed Mode on any tier; they are available only via BYOK on an eligible paid plan, in which case your requests are sent directly to your own Upstream Provider account under your own contract with that provider. We may add, modify, deprecate, or remove supported models, endpoints, or features at any time, including in response to changes by Upstream Providers.

3. Eligibility and Account Registration

You must be at least 18 years old (or the age of legal majority in your jurisdiction) to use the Service. The Service is intended for business and developer use and is not directed to children.

You must create an account to use the Service. Authentication is provided through our identity provider, Clerk. You are responsible for:

  • Providing accurate and complete registration information;
  • Maintaining the confidentiality of your account credentials, Workspace API keys, and any BYOK credentials;
  • All activity that occurs under your account or Workspace, whether or not authorized by you.

You must notify us promptly at contact@proxyllm.dev of any unauthorized use of your account or any suspected security breach. We are not liable for any loss arising from unauthorized use of your account that results from your failure to safeguard your credentials.

4. API Keys and Security

The Service issues you one or more Workspace API keys (prefixed pl_) to authenticate your requests. You are solely responsible for keeping these keys secret. Treat them as you would a password; do not embed them in client-side code, public repositories, or shared environments. Workspace API keys are stored to authenticate your requests and are compared using timing-safe comparison; you can rotate them at any time from the dashboard, which immediately invalidates the prior key.

If you use BYOK (available on eligible paid plans), you may configure your own Upstream Provider API keys within the Service. We encrypt BYOK credentials at rest using AES-256-GCM authenticated encryption with a server-held master key, and we transmit them only over TLS-encrypted connections to the provider or base URL you configure. The cleartext value of a BYOK key is never returned through our API; only a masked fingerprint (last four characters) and the configured base URL are surfaced. When you delete a BYOK credential or close your account, the stored ciphertext is removed. You remain solely responsible for the security, validity, billing, quotas, and acceptable use of your own Upstream Provider keys, and for compliance with the Upstream Provider's terms that govern them.

We implement reasonable technical and organizational security measures, including TLS in transit, network-isolated databases and cache, timing-safe comparison of API keys, and a hardened, non-root runtime container. No method of transmission or storage is completely secure, and we cannot guarantee absolute security.

5. Customer Content, Caching, and Embeddings

You retain all rights, title, and interest in your Inputs, and as between you and ProxyLLM, you own the Outputs to the extent permitted by the applicable Upstream Provider's terms. We claim no ownership of your Customer Content.

You grant us a limited, non-exclusive, worldwide, royalty-free license to host, store, process, transmit, cache, and create derived representations of your Customer Content solely to the extent necessary to operate and provide the Service to you — namely to proxy your requests to Upstream Providers, to perform semantic and exact-match caching, to perform smart routing, and to generate usage and cost analytics. This license terminates when the relevant Customer Content is deleted from the Service or expires from the cache, except as needed to comply with law.

You should understand precisely how the Service handles your data:

  • Request logs store metadata only. Our request logs record the model used, input and output token counts, estimated cost, latency, cache hit/miss status, an optional customer-supplied tag, a request identifier, and a timestamp. Request logs do not contain the text of your prompts or the text of model responses.
  • The semantic cache stores response content and a hashed representation of your prompt. To serve cached results, we store in our cache (Redis): (a) the full text of the corresponding model response as JSON; (b) a SHA-256 hash of the normalized prompt text (specifically the system prompt plus the last user message) used as the cache key — the cleartext prompt is not persisted; and (c) a numerical embedding vector derived from that prompt text, used for semantic matching. Cache entries are scoped to your Workspace and expire automatically according to your plan tier (see Section 8). Cache values are stored as plaintext JSON within our cache, protected by infrastructure-level controls and access restrictions; they are not additionally encrypted at the application layer.
  • Prompt text is sent to an embedding provider. To enable semantic caching, the normalized prompt text (system prompt plus last user message) is transmitted to a third-party embedding provider to generate the embedding vector. This text leaves our infrastructure for that purpose.
  • Semantic matching may return a cached response generated for a similar earlier request. When a new prompt is sufficiently similar to a previously cached prompt within your Workspace, the Service may return the previously cached response instead of calling the Upstream Provider. You are responsible for determining whether semantic caching is appropriate for your use case and for disabling or bypassing it where exact, fresh responses are required.

You are responsible for the lawfulness of your Customer Content and for obtaining all rights, consents, and permissions necessary to transmit it through the Service and to the Upstream Providers.

6. AI Outputs — No Warranty; Your Responsibility

Outputs are generated by third-party AI models and are provided "AS IS". AI models are probabilistic and non-deterministic. Outputs may be inaccurate, incomplete, outdated, biased, offensive, or otherwise objectionable, and may contain content that does not reflect real facts ("hallucinations"). Identical or similar Inputs may produce different Outputs.

We make no warranty regarding the accuracy, reliability, completeness, fitness, non-infringement, or appropriateness of any Output. You are solely responsible for evaluating, reviewing, and validating all Outputs before relying on or acting on them. You must not use Outputs as a substitute for professional advice, and you must not rely on Outputs for medical, legal, financial, safety-critical, employment, credit, or other high-stakes decisions without qualified human review. You assume all risk arising from your use of Outputs.

7. Acceptable Use Policy

You agree not to use the Service, and not to permit any person using your account to use the Service, to:

  • Violate any applicable law or regulation, or infringe the intellectual property, privacy, or other rights of any third party;
  • Attempt to gain unauthorized access to the Service, other Workspaces, or any underlying infrastructure, or probe, scan, or test the vulnerability of the Service without our prior written consent;
  • Interfere with, disrupt, overload, or degrade the Service or the experience of other users;
  • Circumvent, disable, or attempt to exceed the rate limits, quotas, model whitelist, or other restrictions of your subscription plan;
  • Resell, sublicense, or redistribute access to the Service except as expressly permitted in writing;
  • Reverse engineer, decompile, or attempt to extract source code from the Service, except to the extent such restriction is prohibited by law;
  • Use the Service to develop a competing product by copying its features or functionality.

Flow-down of Upstream Provider policies. Because the Service forwards your requests to Upstream Providers, you must also comply with the usage policies of the relevant Upstream Provider, including the OpenAI Usage Policies and the Anthropic Acceptable Use Policy. Without limitation, you must not use the Service to generate, transmit, or facilitate:

  • Child sexual abuse material or any content that sexually exploits minors;
  • Malware, ransomware, or code intended to disrupt, damage, or gain unauthorized access to systems;
  • Content that promotes or enables violence, terrorism, weapons development, or self-harm;
  • Harassment, hate speech, defamation, or unlawful discrimination;
  • Fraud, deception, spam, or impersonation;
  • Automated decision-making that produces legal or similarly significant effects on individuals without appropriate human oversight and required disclosures.

A violation of any applicable Upstream Provider policy is a violation of these Terms. We may suspend or terminate your access, with or without notice, for any actual or suspected violation of this Section, and we may be required to suspend or terminate access at the direction of an Upstream Provider.

8. Plans, Billing, and Subscriptions

The Service offers Free, Pro, and Scale tiers. Plan features, rate limits, the Managed Mode model whitelist, BYOK eligibility, request-log retention, and cache time-to-live differ by tier:

  • Request-log retention: Free 7 days, Pro 30 days, Scale 90 days, after which logs are automatically deleted.
  • Cache time-to-live: Free 24 hours, Pro 72 hours, Scale 168 hours (7 days), after which cache entries automatically expire.

Billing for paid plans is handled by our payment processor, Stripe. We do not store your full card number or payment credentials; Stripe processes them. By subscribing to a paid plan, you authorize us (through Stripe) to charge your payment method on a recurring monthly basis until you cancel. Fees are stated in US dollars, are exclusive of any applicable taxes, and are non-refundable except where required by law or expressly stated otherwise.

You may cancel at any time through the billing portal. Cancellation takes effect at the end of the then-current billing period, and you retain access until that time. We may change plan pricing or features on prospective notice; material changes to pricing will be communicated before they take effect, and continued use after the effective date constitutes acceptance.

Usage and cost figures shown in the dashboard are estimates for your convenience and may differ from amounts billed by an Upstream Provider, particularly under BYOK. They do not constitute an invoice from any Upstream Provider.

9. Service Availability, Upstream Dependencies, and Subprocessors

We strive to maintain high availability but do not warrant that the Service will be uninterrupted, timely, secure, or error-free. The Service may be unavailable during maintenance windows, updates, or events beyond our reasonable control.

The Service depends on third-party infrastructure providers, Subprocessors, and Upstream Providers to function. These currently include, without limitation:

  • Upstream Providers: OpenAI, OpenRouter, and Anthropic (AI model and embedding providers).
  • Infrastructure and platform Subprocessors: Railway (proxy hosting, databases, and cache), Vercel (dashboard and landing-site hosting, and web analytics), Clerk (authentication and identity), and Stripe (payment processing).
  • Operational Subprocessors: Resend (transactional and account email delivery, which receives your email address and the content of those messages), Sentry (application error and performance monitoring), and UptimeRobot (uptime monitoring and the public status page).

This list is provided for transparency and is not exhaustive. The authoritative, current, and dated list of all Subprocessors — including each one's name, processing purpose, categories of data received, and location — is maintained at /subprocessors and is incorporated into these Terms by reference. The categories of personal data processed by each Subprocessor are further described in our Privacy Policy.

We are not responsible or liable for any act, omission, outage, latency, content-moderation decision, model change, deprecation, suspension, rate limit, quota, or price change of any Upstream Provider, Subprocessor, or third-party infrastructure provider, nor for any consequence to you arising from them, including where an Upstream Provider suspends or terminates access to a model or to your (or our) provider account. We may modify, throttle, or remove access to any model or Upstream Provider integration at any time.

Force Majeure. Neither party will be liable for, nor be considered in breach of these Terms on account of, any delay or failure to perform (other than your payment obligations) resulting from causes beyond its reasonable control. Such causes include, without limitation: acts of God, natural disasters, fire, flood, epidemic or pandemic; war, terrorism, civil unrest, or government action, embargo, sanction, or change in law; labor disputes; failures or outages of electricity, internet, or telecommunications networks; denial-of-service, ransomware, or other malicious attacks; and any outage, suspension, rate limit, quota, model deprecation, content-moderation decision, or other act or omission of an Upstream Provider, Subprocessor, or other third-party infrastructure provider. The affected party will use reasonable efforts to mitigate the effect of the event and resume performance. This Force Majeure provision is in addition to, and does not limit, the disclaimers and limitations of liability in Sections 12 and 13, which apply to any such event.

10. Data Protection and Roles

For data protection purposes, the parties acknowledge the following allocation of roles:

  • For account, authentication, billing, and Service-analytics data (such as your email, name, user identifier, Stripe customer and subscription identifiers, and usage metadata), ProxyLLM acts as the controller (in Brazilian terms, controlador).
  • For Customer Content (Inputs and Outputs) that you transmit through the Service, ProxyLLM acts as an operator/processor (operador) processing such content on your behalf and on your documented instructions, which are to provide the Service. In that context, you are the controller of your own end-users' personal data, and the Upstream Providers and our other Subprocessors act as sub-operators/sub-processors.

We will process Customer Content only as necessary to provide the Service and as instructed by you, and we will not use Customer Content for our own purposes. We do not use your Inputs, Outputs, or cached content to train, fine-tune, or improve any AI model. To our knowledge and as of the last-updated date, each of OpenAI and Anthropic excludes API data from model training by default; their then-current terms govern that behavior and may change, and we make no warranty as to their practices. Under BYOK, your traffic is governed additionally by your own contract and opt-out settings with the relevant provider.

Subprocessors and change notification. You grant ProxyLLM general written authorization to engage the Subprocessors listed at /subprocessors to process Customer Content and personal data in connection with the Service. Before engaging a new or replacement Subprocessor, we will update the dated /subprocessors page and notify account owners by email. You may object to a new or replacement Subprocessor on reasonable, good-faith data-protection grounds within fourteen (14) days of that notice; if you do, the parties will work in good faith to resolve the objection, and if it cannot be resolved, your sole remedy is to terminate the affected portion of the Service and discontinue use.

Where you process personal data through the Service in a manner that requires a data processing agreement, a Data Processing Addendum (available on request at contact@proxyllm.dev) applies and is incorporated into these Terms upon execution. Our use of Subprocessors, and the categories of data each receives, are described in our Subprocessors list and Privacy Policy.

Sensitive and special-category data. The Service is not designed to process sensitive personal data (such as data revealing racial or ethnic origin, health, biometric or genetic data, religious or political beliefs, or sexual orientation). You should not submit such data through the Service unless strictly necessary for your use case, and you are responsible for the lawfulness of doing so. Because the semantic cache may store and reproduce Output content, you should avoid placing sensitive data in prompts where that is not appropriate.

Our processing of personal data is described in detail in our Privacy Policy, which addresses your rights under Brazil's Lei Geral de Proteção de Dados (LGPD), the EU General Data Protection Regulation (GDPR), and the California Consumer Privacy Act (CCPA/CPRA), as applicable.

11. Intellectual Property

The Service, including its software, design, documentation, trademarks, and all related intellectual property (excluding Customer Content), is owned by Sysdev TechStrategy & Consulting or its licensors and is protected by applicable law. Subject to your compliance with these Terms, we grant you a limited, non-exclusive, non-transferable, revocable license to access and use the Service for your internal business purposes during the term of your subscription. No rights are granted to you except as expressly set out in these Terms.

If you provide us with feedback or suggestions about the Service, you grant us a perpetual, irrevocable, royalty-free license to use that feedback without restriction or obligation to you.

12. Disclaimer of Warranties

THE SERVICE AND ALL OUTPUTS ARE PROVIDED "AS IS" AND "AS AVAILABLE", WITHOUT WARRANTIES OF ANY KIND, WHETHER EXPRESS, IMPLIED, OR STATUTORY. To the maximum extent permitted by applicable law, we disclaim all warranties, including any implied warranties of merchantability, fitness for a particular purpose, title, non-infringement, accuracy, and any warranty arising from course of dealing or usage of trade. We do not warrant that the Service will meet your requirements, that caching or routing will produce any particular cost saving or result, or that any Output will be accurate or fit for any purpose. Nothing in these Terms excludes or limits any warranty or right that cannot lawfully be excluded or limited, including mandatory consumer protections under the Brazilian Consumer Defense Code (Código de Defesa do Consumidor) where applicable.

13. Limitation of Liability

TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW:

  • In no event will ProxyLLM, Sysdev TechStrategy & Consulting, or its officers, employees, or suppliers be liable for any indirect, incidental, special, consequential, exemplary, or punitive damages, or for any loss of profits, revenue, data, goodwill, or business opportunity, or for the cost of substitute services, arising out of or relating to the Service or these Terms, whether based in contract, tort (including negligence), strict liability, or otherwise, and even if advised of the possibility of such damages.
  • We will not be liable for any damages arising from (a) any Output, including its inaccuracy or your reliance on it; (b) any act, omission, outage, change, or decision of an Upstream Provider, Subprocessor, or third-party infrastructure provider, or any event of Force Majeure described in Section 9; (c) loss or corruption of Customer Content, including the expiry or clearing of cache entries; or (d) unauthorized access to your account resulting from your failure to safeguard credentials.
  • Our total aggregate liability arising out of or relating to the Service or these Terms will not exceed the greater of (i) the total fees you paid to us for the Service in the twelve (12) months immediately preceding the event giving rise to the liability, or (ii) one hundred US dollars (US$100).

These limitations apply even if a remedy fails of its essential purpose. Nothing in this Section limits liability that cannot be limited under applicable law, including, where applicable to consumers, mandatory protections under the Brazilian Consumer Defense Code.

14. Indemnification

You agree to defend, indemnify, and hold harmless ProxyLLM, Sysdev TechStrategy & Consulting, and its officers, directors, employees, and agents from and against any claims, demands, damages, liabilities, losses, costs, and expenses (including reasonable legal fees) arising out of or related to: (a) your Customer Content; (b) your use of the Service in violation of these Terms or applicable law; (c) your violation of any Upstream Provider's policies or terms; (d) your violation of any third party's rights, including intellectual property or privacy rights; or (e) your failure to comply with your obligations as a controller of your own end-users' personal data. We will provide you with reasonable notice of any such claim and may participate in its defense with counsel of our choosing.

15. Suspension and Termination

You may stop using the Service and close your account at any time. We may suspend or terminate your access to the Service, in whole or in part, with or without notice, if (a) you breach these Terms or the Acceptable Use Policy; (b) we are required to do so by law or by an Upstream Provider; (c) your use poses a security, legal, or operational risk; or (d) for non-payment of fees.

Upon termination, your right to use the Service ceases immediately. We will delete or anonymize your personal data in accordance with our Privacy Policy and applicable law, subject to legal retention obligations and to records held by our payment and authentication processors under their own retention policies. Sections that by their nature should survive termination — including Sections 5, 6, 8 (with respect to fees and payment obligations accrued before termination), 9, 10 (with respect to data-protection roles and our no-training commitment), 11, 12, 13, 14, 16, and 17 — will survive.

16. Changes to the Service and to These Terms

We may modify, suspend, or discontinue any part of the Service at any time. We may also update these Terms from time to time. We will notify you of material changes by email or through the dashboard before they take effect where practicable. Changes take effect on the stated effective date, and your continued use of the Service after that date constitutes acceptance of the updated Terms. If you do not agree to the updated Terms, you must stop using the Service.

17. Governing Law and Jurisdiction

These Terms are governed by and construed in accordance with the laws of the Federative Republic of Brazil, without regard to conflict-of-laws principles. Except where mandatory law (including consumer-protection law) provides otherwise, the parties elect the courts of the domicile of Sysdev TechStrategy & Consulting in Brazil (as stated in Section 18) as the exclusive venue to resolve any dispute arising out of or relating to these Terms or the Service. Nothing in this Section deprives a consumer of the protection of mandatory provisions of the law of the consumer's place of residence, including, for Brazilian consumers, the right to bring suit at their own domicile under the Brazilian Consumer Defense Code.

If any provision of these Terms is held unenforceable, that provision will be limited or severed to the minimum extent necessary, and the remaining provisions will remain in full force and effect. Our failure to enforce any right or provision is not a waiver of that right or provision. You may not assign these Terms without our prior written consent; we may assign them in connection with a merger, acquisition, or sale of assets.

18. Contact

The Service is operated by:

Sysdev TechStrategy & Consulting Ltda. CNPJ: 37.016.893/0001-73 Av. Andrômeda, 433, Sala 515, Jardim Satélite, São José dos Campos – SP, CEP 12.230-000, Brazil Email: contact@proxyllm.dev

For questions about these Terms, to exercise data-protection rights, or for any other inquiry, contact us at the email above. Our data-protection representative (encarregado/DPO) and channels for exercising your rights are identified in our Privacy Policy.