LEGAL & PLATFORM DISCLAIMERS

Disclaimer

Effective Date: September 1, 2026 • Last Updated: September 3, 2026 • Scope: AI Inference & Proxy Services

Disclaimer Summary: AgentMink is a proxy routing and cost arbitrage gateway infrastructure platform. We do not author or generate AI model outputs. Artificial intelligence outputs are probabilistic and may contain inaccuracies or hallucinations. Users must independently test and validate all AI outputs before production deployment.

1. Role as an Infrastructure Gateway Provider

AgentMink operates exclusively as an intermediary proxy routing and cost arbitrage infrastructure service. AgentMink does not create, own, train, or author the weights, safety filters, or generative algorithms of connected third-party models (such as OpenAI, Anthropic, Google Gemini, Groq, DeepSeek, Mistral, Together AI, or AWS Bedrock).

All inferences, completions, embeddings, code generations, and responses delivered through our proxy gateway originate directly from the upstream third-party model providers selected by your configured routing strategy or request parameters.

2. Nature of Artificial Intelligence & Output Accuracy

Generative artificial intelligence and Large Language Models (LLMs) are complex, probabilistic systems. By utilizing AgentMink, you acknowledge and agree that:

  • Potential Inaccuracies & Hallucinations: AI models may generate responses that are factually incorrect, misleading, incomplete, biased, or non-functional.
  • No Verification by Gateway: AgentMink is a stateless proxy that does not inspect, modify, fact-check, or validate the substantive accuracy of AI prompt responses transmitted through the network.
  • Customer Verification Requirement: You bear sole responsibility for evaluating, testing, human-reviewing, and validating all model-generated code, text, documents, and recommendations prior to deploying them in production, customer-facing, or mission-critical applications.

3. No Professional, Legal, Medical, or Financial Advice

No content, documentation, benchmark comparison, or model output routed through AgentMink should be construed as professional advice:

  • No Legal Advice: Model outputs do not constitute legal counsel, contract review, or compliance certification.
  • No Medical Advice: Model outputs must not be used as clinical diagnostic tools or substitutes for professional medical evaluation.
  • No Financial Advice: Model responses do not constitute investment, tax, financial, or accounting guidance.
  • No Engineering Warranty: Automatically generated software code or infrastructure scripts must be independently reviewed and vulnerability-scanned prior to execution.

4. Third-Party Model Providers & Upstream Outages

While AgentMink incorporates automated failover mechanisms to catch HTTP 429, 500, 502, 503, and 504 errors in under 50ms, we do not control the uptime, infrastructure availability, capacity, or performance of upstream model providers.

AgentMink is not liable for damages, service interruptions, or financial losses resulting from:

  • Global or regional outages of third-party model APIs (e.g., OpenAI, Anthropic, Google Cloud service interruptions).
  • Sudden deprecation, retirement, or parameter modifications of upstream model endpoints.
  • Upstream rate-limit throttling, context-length reductions, or account suspensions imposed directly by third-party vendors.
  • Upstream token pricing modifications or adjustments enacted by model providers.

5. Autonomous Routing Strategies & Dynamic Cost Arbitrage

AgentMink provides dynamic routing algorithms (cost_first, quality_first, latency_first) that select upstream models based on continuous edge latency monitoring and pricing matrices.

  • Pricing Matrix Freshness: Multi-provider token pricing is synchronized automatically every 6 hours. While we strive for absolute pricing accuracy, transient price discrepancies resulting from unannounced vendor rate adjustments may occasionally occur.
  • Savings Estimates: Cost reduction figures and arbitrage percentages presented on our website (agentmink.com) and dashboard (app.agentmink.com) are illustrative estimates based on standardized benchmark token consumption and historical telemetry. Actual savings vary based on prompt lengths, completion tokens, and model choices.

6. “AS IS” and “AS AVAILABLE” Disclaimer of Warranties

TO THE MAXIMUM EXTENT PERMITTED UNDER APPLICABLE LAW, THE AGENTMINK SERVICES, PROXY GATEWAY ENDPOINTS, DOCUMENTATION, AND DASHBOARD PLATFORM ARE PROVIDED STRICTLY ON AN “AS IS” AND “AS AVAILABLE” BASIS. AGENTMINK EXPRESSLY DISCLAIMS ALL WARRANTIES, REPRESENTATIONS, AND CONDITIONS OF ANY KIND, WHETHER EXPRESS, IMPLIED, STATUTORY, OR OTHERWISE, INCLUDING WITHOUT LIMITATION WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, TITLE, NON-INFRINGEMENT, AND UNINTERRUPTED OR ERROR-FREE OPERATION.

7. External Links & Third-Party Integrations

Our documentation (agentmink.com/docs) and integrations catalog (agentmink.com/integrations) may reference third-party open-source libraries, developer frameworks (such as Vercel AI SDK, LangChain, LlamaIndex, AutoGen, CrewAI), and external documentation.

AgentMink does not endorse, control, or assume responsibility for the stability, licensing, security, or maintenance of any third-party software libraries, packages, or external websites.

8. Contact Information

If you have questions or require clarification regarding this Disclaimer, please contact our legal and support teams: