Full-strength Claude and GPT
One base_url away
Same models, same versions, nothing downgraded. Compatible with the official SDKs — change one line and you are running.
$1 of list usage billed at ¥7 · per-token · credit never expires
7 models — see full pricing →Claude Code · Codex · Cline · Cursor · Python / Node SDK all work as-is
See what you spent, and where
Balance, today's spend and the last 30 days on one screen. Every request records its model, token counts and charge — filter by key or by date.
Common questions
How is this different from calling the vendors directly?
The wire protocol and response shape are identical — tool calls, streaming and system prompts all behave the same. What differs is billing: you pay in RMB without needing an overseas payment method, and the same usage typically costs considerably less than list price in USD.
How does billing work? Any monthly fee or minimum?
You pay for the input and output tokens each request actually uses. No monthly fee, no tiered bundles, no concurrency charges, no minimum spend. Credit is held in RMB — what you add is what you get, and top-ups land immediately.
Does credit expire?
No. Credit has no expiry date and sitting idle costs nothing.
Can I audit every charge?
Yes. The console lists each request with its model, token counts and amount charged, filterable by key or date, so it reconciles line by line against your statement.
Which clients are supported?
Anything that lets you set a custom base_url — the official Python and Node SDKs, Claude Code, Codex, Cline, Cursor and others.
Ready when you are
Create a key as soon as you sign up. Add a small amount, confirm it works, then decide.
- 01
Sign up
Email and a verification click. No manual review.
- 02
Create a key
Set a spend cap and expiry per key.
- 03
Swap base_url
Into your env vars or SDK options.
Synapsea · pay-as-you-go model API access · charges follow live console data