Relay comparison
API relay vs direct provider access
Direct provider access is often the fastest way to start. A relay becomes useful when teams need consistent controls across apps, keys, users, or environments.
Direct provider access
Direct access is simple and has fewer moving parts. It is often enough for experiments, single-user scripts, and teams that already have internal billing controls.
Relay access
A relay adds policy before the upstream request and logs after completion. This makes spend easier to cap, inspect, and allocate.
- Prepaid balance protection
- Per-key model and budget policy
- One request ledger across tools and teammates
When to choose each
Use direct access for minimal prototypes. Use a relay when multiple people, environments, or customer-facing features share AI spend.
FAQ
Does a relay add latency?
A gateway adds a small network hop. Teams usually accept that tradeoff when they need policy and auditability.
Can a relay still use OpenAI-compatible SDKs?
Yes, when the relay exposes compatible base URLs and request formats.