What is AnLink in the current release?
It is a open account access layer for guided testing of a small set of Chinese models with one OpenAI-compatible request surface, console-issued API keys, trial balance, and usage visibility.
FAQ
Questions developers ask before and during first integration.
This page focuses on the current Start. It is written to answer the practical questions that show up during registration, key issuance, model selection, and first request validation.
Answers
Common questions
How do I start using it?
Open https://console.anlinkai.com, register, create an API key, then call https://api.anlinkai.com/api/v1/chat/completions with a currently active model code.
Do I need to create separate upstream accounts for every provider?
Registered users can test through one AnLink base URL first. Separate upstream provider onboarding may still be needed later for direct production contracts.
How does trial balance work?
Up to 2 USDT trial balance: 1 USDT to start, 1 USDT after account setup. Balance and recorded cost are visible in the console.
Can I recharge online right now?
Self-service payment is not part of the current guided testing phase. Top-up bonus: top up with USDT and receive 20% extra usable balance after confirmation.
Where can I compare models and pricing?
Use the model catalog for enabled model routes and Billing and Recharge for automatic TRC20 USDT recharge guidance and manual review rules for exceptions.
Which endpoint should I use first?
Use POST /chat/completions first. That is the primary path the Start is designed around.
Should I expose the API key in frontend code?
No. Keep the key in server-side environment variables or backend secret storage.
Why does a model call fail even though my request shape looks correct?
Check three things first: whether the model is active, whether your account still has USDT balance, and whether the key is still active.
Where can I inspect failures?
Use the Usage Logs page in the console and compare request status, token counts, and any recorded error code.
Does the current Start support streaming?
Yes. Streaming is available through OpenAI-compatible SSE for enabled models such as qwen-flash and deepseek-v4-flash. Usage Logs still record the final token usage and cost when the upstream returns usage metadata.
Can model pricing change?
Yes. During current validation, pricing and enabled model exposure may change while routing and support are still being tuned.
Is this ready for enterprise team workflows?
Not yet. Org controls, advanced routing, broad catalog expansion, and public SLA are not part of the current release.
Support
Need help with developer testing?
Email us with your use case or join Discord for product rollout notes and guided testing support.