One endpoint for every model
A self-hosted gateway in front of multiple providers. One OpenAI-compatible API, per-key budgets and rate limits, request logging and spend tracking.
OpenAI-compatible
Point any SDK at /v1. Chat completions, streaming, tool calling —
same wire format as upstream providers.
Keys with limits
Virtual keys per client, monthly budget, requests per minute and model allow-lists. Spend is attributed per key and per model.
Observability
Request logs, latency and token accounting stored in your own database. Health endpoints for liveness and readiness.