Router documentation

Authentication

Router authenticates with a Bearer token. Pass your key in the Authorization header. The OpenAI SDK sets this for you when you provide api_key.

curl https://router.kyzon.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_KYZON_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai/gpt-4.1-mini",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Making requests

Router implements the OpenAI Chat Completions API. Any request your code already sends works once the base URL points at Router. Below is only what is specific to Router.

# switch providers by changing the slug
model="openai/gpt-4.1-mini"
model="anthropic/claude-sonnet-4"
model="google/gemini-2.5-flash"

stream = client.chat.completions.create(
    model="openai/gpt-4.1-mini",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True,
)

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

Chat completions and streaming are supported with the parameters in this guide. Fine-tuning, assistants, image generation, and file endpoints are not available on Router today.

Supported models

Pass the slug in the model field. This is the current routable set.

Request logs

Every request and response through Router is logged on Australian infrastructure, giving you a full record for debugging and audit.

What is captured

Request and response payloads, model slug, provider, token counts, cost, latency, team, and timestamp.

Where it lives

AWS Sydney (ap-southeast-2).

Retention

Between 30 days and 6 months depending on the use case.

Access

Dashboard export today. Logs API access is available on request during early access.

Errors and rate limits

Router returns standard HTTP status codes with a JSON error object in the OpenAI format: error.message, error.type, and error.code.

Rate limits during early access are set per account when your key is issued. Contact us if you need higher throughput.