JetRouter is an OpenAI-compatible router. Point your existing client at one base URL and call the models in the catalog with the same request shape, one API key and one bill — with the per-token price published before you send anything. Models open for traffic as they are enabled; the Models page shows which are.
The request shape
The same body an OpenAI client already sends, with a key from your dashboard and a catalog slug in model. Use a model the Models page lists as available; models open for traffic as they are enabled.
curl https://jetrouter.dev/v1/chat/completions \
-H "Authorization: Bearer $JETROUTER_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "anthropic/claude-sonnet-5",
"messages": [
{ "role": "user", "content": "Summarise this in one sentence." }
]
}'Migration
If your code already speaks to OpenAI, it already speaks to JetRouter. Two lines move, and the model string becomes a catalog slug.
client = OpenAI(
api_key=os.environ["PROVIDER_API_KEY"],
)
client.chat.completions.create(
model="gpt-5",
messages=messages,
)client = OpenAI(
base_url="https://jetrouter.dev/v1",
api_key=os.environ["JETROUTER_API_KEY"],
)
client.chat.completions.create(
model="anthropic/claude-opus-5",
messages=messages,
)Coverage
Every model is a slug in one catalog with one request shape. Adding a vendor to your product stops being an integration and becomes a string.
What the router does
Failover, cost accounting, per-key policy and a cache are the same four problems for every model you add. They are solved once here, on the request path.
Cost
Not a monthly total you reconcile later. Each response carries the charge, the provider that served it and whether the cache paid for it.
X-JetRouter-ProviderX-JetRouter-CacheX-JetRouter-Cost-USDX-JetRouter-Saved-USDJetRouter began as a PII proxy for Hong Kong businesses, and that pipeline is intact. It is now a setting on an API key rather than the product: pii_mode is off, warn or tokenize, and it defaults to off.
Fail closed, by design
Credentials are never readable
No prompt bodies in logs
We hold no SOC 2 or ISO 27001 certification and do not claim one. The statements above describe what the code does and are testable against it.
Pricing
Top up credit and every call draws from it, metered per token in US dollars at the published rate. No subscription and no monthly fee.
one-off deposit
$1.28 of usage
one-off deposit
$12.82 of usage
one-off deposit
$64.10 of usage
one-off deposit
$320.51 of usage
Create an account, mint a key and top up credit. Requests run against the models listed on the Models page as each is opened for traffic. You pay per token from your balance, with no subscription.