Turn upstream model subscriptions into controlled, billable, observable API services with clean limits and no surprises.
$ curl -X POST https://api.rabah.dev/v1/chat/completions \
-H "Authorization: Bearer rk_live_..." \
-d '{"model": "gpt-4", "messages": [...]}'
{
"id": "chatcmpl-abc123",
"usage": { "total_tokens": 42 },
"cost": "$0.00126"
} One gateway. Clear limits. Full visibility.
Track usage, latency, and cost with enough detail to catch drift before it becomes a billing problem.
Issue keys with model permissions, spending caps, and rate limits. One upstream account, many controlled surfaces.
Route OpenAI, Claude, Gemini, and other providers through one stable interface with fallback behavior.
Export clean records by user, model, time window, and project for reconciliation and resale billing.
Add provider credentials for OpenAI, Anthropic, Google, and other upstream AI services.
Create scoped keys, rate limits, model permissions, and spending rules for every downstream user.
Route traffic, monitor usage, and reconcile cost from a single control plane.
Start small, route carefully, and scale once usage is visible.