Skip to content

Developer documentation / Preview

Connect your
first application.

Integration guidance for the development preview. You’ll need an approved pilot or a configured local gateway; this website does not issue API keys.

01 / Connect

Your first request

The examples use https://api.routercore.example/v1 as a placeholder. During an approved pilot, replace it with your assigned endpoint, supply your RouterCore API key, and use a model enabled for your deployment.

API access coming soon. Example endpoint only; approved pilots receive their endpoint and credentials during onboarding.

python
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.routercore.example/v1",
    api_key=os.environ["ROUTERCORE_API_KEY"],
)

resp = client.chat.completions.create(
    model="anthropic/claude-sonnet-4.6",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)

02 / Authenticate

One key for your integration

Send your key as a bearer token. Keys are shown once when created: store the secret securely and use environment variables in your application.

bash
Authorization: Bearer YOUR_ROUTERCORE_KEY

Create and revoke keys in the console during an approved pilot. Request early access to discuss your integration; this public website does not issue API keys.

03 / Choose

Start with a model ID

Use a provider-qualified identifier from the prototype catalog. Catalog metadata describes context windows, capabilities and reference rates; it does not guarantee that a provider is configured or a model is enabled for your key.

text
openai/gpt-5-mini
anthropic/claude-sonnet-4.6
google/gemini-2.5-pro

The gateway’s GET /v1/models endpoint exposes model metadata for your integration. RouterCore virtual models and cross-provider routing remain subject to the product’s engineering milestones.

04 / Handle

Errors you can trace

Errors use a consistent envelope with a status code, type, message, provider and request ID. Keep the X-Request-Id response header when investigating a failed call.

json
{
  "error": {
    "code": 401,
    "type": "authentication_error",
    "message": "Missing API key.",
    "provider": null,
    "request_id": "00000000-0000-4000-8000-000000000001"
  }
}

Illustrative error envelope.

  • 400 — invalid request
  • 401 — missing or invalid key
  • 402 — insufficient credits
  • 429 — configured usage limit reached
  • 502–504 — provider failure or timeout

05 / Control

Make usage intentional

The existing preview includes key and organization spend limits. Review the configuration for your deployment before sending traffic. Broader project budgets, fail-closed policy evaluation and production rate limiting each have their own review gates.

Stopping a response and billing

Before provider dispatch, disconnecting stops the request without a charge. After dispatch, closing a connection or choosing Stop receiving in the Playground only stops delivery. Generation continues within its configured time limit; a completed request is billed from provider-reported usage, including output you did not receive. Set max_tokens or max_completion_tokens to limit generated output, and check Logs and Billing before retrying an interrupted request.

06 / Know the boundaries

A foundation in progress

This documentation describes a development preview. Provider abstraction, policy, usage reconciliation, billing and security must pass the agreed milestones before controlled external production traffic.

Regional data planes, private inference, enterprise identity and intelligent routing are later milestones. The status page is an illustrative design preview, not a live operational feed.