Amanai coordinates access to AI models through changing upstream routes; a model label is not a promise of one permanent provider path.
Amanai's role
Amanai provides a compatibility and routing layer between your client and AI-serving infrastructure. We do not develop, own, or operate every model or upstream service available through the platform.
Routing, compatibility, and fallback
A public model identifier can be mapped to a compatible upstream route. We may select, replace, or retry routes based on availability, capacity, latency, quality, compatibility, abuse risk, and commercial conditions. A fallback may be from a different provider or infrastructure path while preserving the requested API format where practical.
Do not use a model label as proof that a request followed one specific provider, account, geographic path, or infrastructure route unless Amanai explicitly provides that guarantee in writing.
Upstream dependency and change
Upstream services may experience maintenance, overload, rate limits, degraded quality, policy or contract changes, price increases, model retirement, regional restrictions, account suspension, security incidents, or complete service closure. Amanai may adjust pricing, limits, routing, model names, compatibility, or availability; pause sales or access; or retire a model or feature. We will try to communicate material user-facing changes when practical, but emergency changes can occur without advance notice.
Request data flow
The full inference payload needed to answer your request—including prompts, messages, context, tools, images, audio, video, documents, or relevant parameters—is forwarded to upstream infrastructure. Amanai does not add your account email, IP address, user ID, or API key ID to the standard inference payload. The Amanai gateway processes prompt and response content in memory and does not persist ordinary chat content in its database.
Upstream processors may independently log, retain, review, or otherwise process content under their own terms and technical practices. Amanai cannot extend its gateway-level non-retention statement to every upstream. Avoid submitting content whose disclosure to an upstream would be unacceptable.
Uploaded media and generated artifacts
File-based operations require temporary storage so uploads can be validated, processed, and delivered. Files and generated artifacts are bound to the authenticated owner; possession of another user's identifier does not grant access. Download links can be short-lived and should not be treated as permanent public hosting. Delete uploaded files when they are no longer needed and copy generated artifacts you intend to keep before their delivery link expires.
Output limitations
AI output can be inaccurate, incomplete, outdated, offensive, biased, insecure, or fabricated. Similar requests can produce different results. Amanai does not warrant factual accuracy, fitness for a specific purpose, originality, non-infringement, or continuity of any output. You are responsible for review, testing, attribution, licensing checks, and human oversight appropriate to your use case.
Do not rely on output as the sole basis for medical, legal, financial, employment, safety-critical, emergency, or other high-impact decisions.
Sensitive and regulated data
Do not submit passwords, authentication secrets, private keys, seed phrases, payment credentials, government identifiers, confidential production data, health records, or other highly sensitive or regulated information unless you have independently determined that the full processing path is appropriate and lawful. Amanai is not offered as a dedicated regulated-data environment and does not enter into a DPA, BAA, or sector-specific compliance commitment unless separately agreed in writing.
Third-party names and brands
Claude, Anthropic, OpenAI, GPT, GitHub, Google, Grok, xAI, Qwen, and other names and marks belong to their respective owners. References identify compatibility or model access and do not imply sponsorship, endorsement, partnership, or affiliation.