FAQ
What is Accly Gateway?
Accly Gateway routes API requests to supported third-party AI models. Accly provides the account, key, model catalog, compatible endpoints, usage controls, and billing layer; the model provider defines model behavior. See Gateway API.
How do I send a first request?
Create a paid account API key and send a chat request. Follow Quickstart for the exact request.
Which API formats are supported?
Supported formats include OpenAI-compatible Chat Completions and Responses, Anthropic Messages, and native Google Gemini routes. Embeddings, reranking, and audio use additional model-specific endpoints. A model must support the endpoint you select; see Gateway API and Models.
Do I need a provider API key?
No. Send your Accly API key to Accly endpoints. Never send a provider key to the Gateway. Keep your Accly key in a server-side secret store; see Authentication.
Can I use any model with any endpoint?
No. Check the model’s supported protocols and capabilities in the Model Hub. Provider model IDs, prompts, parameter semantics, and modality limits are not interchangeable. See Models for provider references.
How do audio models work?
Accly currently exposes OpenAI audio through Chat Completions, Google TTS through native Gemini generateContent, and MiniMax TTS through native T2A routes. Their payloads and output formats differ. See Audio & Speech.
What are embeddings and reranking?
Embedding endpoints produce vectors; reranking scores a supplied set of documents against a query. They do not generate answer text. See Embeddings & Reranking.
Why was my API request blocked?
The error response includes a code that identifies the failure class. Check the model/plan access, key scope, usage state, request format, and Retry-After where present. See Errors & Limits for the code reference.
Are prompts sent to model providers?
Yes. Prompts and uploaded content can be sent to the selected third-party provider to generate a response. Read AI Safety & Data and the linked legal terms for storage and training details.
Does Accly train its own models on my requests?
Accly does not use customer chats or Gateway API requests to train Accly-owned foundation models. Provider processing still applies; see AI Safety & Data.
What do request units mean?
Request units are weighted plan usage capacity, not cash or a one-request/one-unit balance. The calculation and reset behavior are in Pricing & Usage, and exact quotas are in Plans.
Can I buy more usage after the daily allowance?
Eligible paid accounts can enable prepaid fallback. It does not create API-key access or bypass model and rate limits. See Prepaid Credits.
Where can I get billing help?
Manage subscriptions in the Accly app. For checkout, cancellations, or payment failures, see Billing; for exact plan amounts, see Plans. Refund requests follow the Refund & Cancellation Policy .