One endpoint.
Every model worth calling.
Frontier and open-weight models behind a single API — routed by task, backed up automatically when a provider stumbles, capped by a budget you set. Access is by request while we scale it carefully.
Tell us what you’re building. A person reviews it and emails you a code.
One gateway. It decides who answers.
You send a request and a routing policy — not a provider name. The gateway picks a model, and tells you why.
POST /v1/route
{
"input": "…",
"route": {
"prefer": ["frontier-a", "frontier-b"],
"fallback": ["open-weight-a"],
"budget_usd": 0.02,
"quality_floor": 0.90
}
}
200 OK
{
"served_by": "frontier-b",
"reason": "frontier-a over budget at current load",
"cost_usd": 0.011,
"quality_score": 0.94
}Four decisions the gateway makes for you.
One endpoint
Call a single URL. The gateway decides which model actually handles the request — you never hardcode a provider.
Automatic fallback
A provider errors, rate-limits, or degrades — the next model in your policy picks it up mid-request, not after you notice.
Budget ceilings
Set a hard budget per call or per policy. The gateway routes around anything that would blow it, instead of just billing you for it.
Quality-based routing
Route on a quality floor, not just price — cheap models take what they’re good at, frontier models pick up the rest.
No self-serve signup. On purpose.
We’re early. Access is granted by request, reviewed by a person, not issued automatically.
- 01
Tell us what you’re building
Name, email, a link to the project, and how long you need access for.
- 02
A person reviews it
Not a bot — someone actually looks at the link before anything is issued.
- 03
Your code lands by email
Sent from support@plasmient.com the moment it’s approved.
Have something that needs routing?
Tell us what you’re building and we’ll get you a code.