One endpoint · two providers

The reliable control plane between your product and AI.

Route each task to the right model, fail over safely, guarantee JSON, and keep a complete receipt for every request. Then benchmark new models against the real work they would replace.

Start routing Sign in to Router

Route failover ready

Task-aware model control

Set a default route, then override it by service, task, or identity without changing client code.

sense → call_tags → gemini-2.5-flash

Observe JSON valid

Every call leaves a receipt

Inspect inputs, outputs, attempts, latency, tokens, exact spend, JSON repair, and the route decision behind it.

Attempts: 1
Latency: 1.8s
Status: ok

Benchmark real traffic

Compare before you switch

Replay past requests across models and compare structure, assertions, reliability, speed, and cost with deterministic evidence.

baseline$0.42

candidate$0.19

Compatible by design

Keep the familiar chat completions contract.

Swap the base URL and API key, then add Router context when you need it. OpenRouter stays first; Requesty is ready when a route or provider needs protection.

POST /v1/chat/completions