About Relay

Why we exist

Open and affordable models are now good enough for most production work, but using them well is annoying: prices differ across providers, the same model is served at very different rates, and most of the real savings (caching, batching, smart routing) are left on the table. Meanwhile, gray "relay" services quietly downgrade models or hide their markup.

Relay exists to give builders the lowest effective cost — not just a low sticker price — through one OpenAI-compatible API, with complete transparency and no silent downgrades.

What we believe

• Effective cost > sticker price. Caching, batch and routing should be on by default and visible in your bill.
• Transparency wins. You should see which provider served each call and exactly what it cost.
• No silent downgrades. You get the model you asked for, verifiably.
• Own your stack. Open tools, no lock-in. For teams that want more, we also build and run custom agents.

How we make money

A small, visible fee on top of the cheapest provider price — never by inflating token rates. See pricing.

Get in touch

Questions, partnerships, or press: contact us.

Relay is an independent service and is not affiliated with the model or provider brands referenced on this site.