Introduction

What the gateway does

Every request arrives at one endpoint, is authenticated against your API key, priced against your balance, and routed to a provider that can serve the model you named. If that provider rate limits or fails, the request is retried against the next one before an error ever reaches you.

• One API surface. The request body, the response body, the streaming format and the error envelope all match OpenAI.
• One balance. Credit is drawn down per request across every provider, so there is no per-vendor account to top up.
• One log. Model, provider, tokens, latency and exact cost are recorded for every call.

The shortest possible example

Point the base URL at Openbase, pass an Openbase key, and name a model by its provider-prefixed slug.

curl https://api.openbase.ai/v1/chat/completions \
  -H "Authorization: Bearer $OPENBASE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-sonnet-5",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Where to go next