One endpoint. Every model.

Introduction

JTRouter is a unified AI gateway that exposes a single OpenAI-compatible endpoint for every major language model. You write one integration — we handle model selection, fallback, and billing. No additional SDKs, no per-provider auth flows, no version drift.

The API is a strict superset of the OpenAI Chat Completions format. Any existing code that calls OpenAI can point at JTRouter with two changes: swap the base URL and add your JTRouter key.

Quickstart

Grab your API key from the Keys dashboard and send your first request. Set model to auto to let JTRouter pick the optimal model automatically.

curl https://api.jtrouter.io/v1/chat/completions \
  -H "Authorization: Bearer $JTROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [
      { "role": "user", "content": "Explain monads in plain English." }
    ]
  }'

The response is identical to the OpenAI format, with a few extra headers that tell you which model was selected, the actual latency, and the exact cost charged.

HTTP/2 200
x-jt-model: claude-sonnet-4-6
x-jt-route: quality
x-jt-latency-ms: 1124
x-jt-cost-usd: 0.00031
x-jt-fallback: false

Authentication

All requests must include a bearer token in the Authorization header. Keys are prefixed jtk_ and are scoped to your workspace. You can create multiple keys with different rate limits and permission scopes from the dashboard.

Routing

When model is set to auto, JTRouter scores every available model against four dimensions — quality, latency, cost, and context fit — then routes to the highest scorer above your configured quality floor. Scoring runs in under 2 ms and is invisible to callers.

You can bias the routing strategy with the x-jt-route request header:

curl https://api.jtrouter.io/v1/chat/completions \
  -H "Authorization: Bearer $JTROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -H "x-jt-route: latency" \
  -d '{
    "model": "auto",
    "messages": [{ "role": "user", "content": "Hello" }]
  }'

Fallback

If the selected model returns a non-200 status, JTRouter automatically retries with the next-best candidate from the scoring list. This happens synchronously — from your perspective the request either succeeds or returns a single error after all candidates are exhausted. The x-jt-fallback response header indicates whether a fallback occurred.

Models

Pass any JTRouter alias as the model field to pin a specific model. The full list — with live availability, context windows, and per-token pricing — is always up to date in the Models catalogue.

To pin to Claude Sonnet specifically, replace auto with claude-sonnet:

curl https://api.jtrouter.io/v1/chat/completions \
  -H "Authorization: Bearer $JTROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet",
    "messages": [
      { "role": "user", "content": "Write a unit test for this function." }
    ]
  }'

Errors

JTRouter uses standard HTTP status codes. All errors return a JSON body with error.code, error.message, and an optional error.param field identifying the offending input field.