OpenAI-compatible AI gateway
One endpoint. Every frontier model.
Point your existing OpenAI SDK at a single base URL and reach GPT, Claude, Gemini, DeepSeek and 100+ models with one key. Prepaid credits, no monthly fee, no markup on inference.
No card needed to start. Credits never expire.
from openai import OpenAI
client = OpenAI(
base_url="https://api.bestokens.cn/v1",
api_key="sk-bs-...",
)
resp = client.chat.completions.create(
model="gpt-5-mini",
messages=[{"role": "user", "content": "Hello"}],
)Built for people shipping, not for demos
The boring parts of running AI in production, handled.
Drop-in compatible
Change base_url and api_key. Your existing OpenAI SDK code, tools and evaluation harnesses keep working unchanged.
No inference markup
You pay the upstream model price. Our only revenue is a platform fee on top-up, so heavy usage stays cheap.
Edge-routed worldwide
Requests land on the Cloudflare edge closest to you and fail over across channels the moment a provider degrades.
Metered to the micro-dollar
Integer micro-USD accounting behind a per-user Durable Object, so concurrent requests can never overspend your balance.
Live in three steps
Under five minutes from sign-up to first response.
Create a key
Sign up with an email address, open the dashboard and create your first key. The full key is shown once.
Top up credits
Pay by card through Creem or with crypto. Credits land in your balance instantly and never expire.
Swap the base URL
Set base_url to https://api.bestokens.cn/v1. Nothing else in your code changes.
Works with the tools you already use
Anything that speaks the OpenAI protocol works out of the box: LangChain, LlamaIndex, Vercel AI SDK, Dify, n8n, Cursor, Continue and more.
- OpenAI SDK
- LangChain
- LlamaIndex
- Vercel AI SDK
- Dify
- n8n
- Cursor
- Continue
- SillyTavern
- Open WebUI
# any OpenAI-compatible client
export OPENAI_BASE_URL="https://api.bestokens.cn/v1"
export OPENAI_API_KEY="sk-bs-..." A small fee on top-up, never on tokens
Compare with a 5.5% top-up fee elsewhere. The more you commit, the lower the percentage.
View pricing| Top-up | Platform fee | Credited |
|---|---|---|
| $2,000+ | 2.5% | $1950.00 |
| $500 – $1,999 | 3% | $485.00 |
| $100 – $499 | 3.5% | $96.50 |
| $5 – $99 | 4% | $19.20 |
Questions people ask first
How is this different from calling OpenAI directly?
Same models, same request format. You get one key across every provider, one bill, automatic failover, and no need for separate accounts or cards per vendor. We charge a platform fee on top-up instead of marking up tokens.
Do you charge more than the official model price?
No. Inference is billed at the upstream list price. Our revenue is the 2.5%–4% platform fee applied when you top up.
What happens when my credits run out?
Requests are rejected with HTTP 402 before they reach the provider, so you are never charged for a call you cannot afford. Top up and the same key works again immediately.
Can I get an invoice for my company?
Yes. Above $2,000 per month we issue invoices and accept wire transfer. Contact us with your expected volume.
Is my prompt data used for training?
No. Prompt and completion bodies are never stored. We keep token counts, model id, latency and status for 30 days for billing and abuse detection.
Which payment methods do you accept?
Cards and wallets through Creem — Visa, Mastercard, Amex, Apple Pay, Google Pay, Alipay — plus crypto through NOWPayments covering BTC, ETH, USDT, USDC and 300+ assets.
Ship your AI feature this week
Create a key, top up $5 and make your first call in under five minutes.