About shuntio

What we do

• Serve the OpenAI and Anthropic wire protocols from one endpoint, so an existing SDK works after changing a base URL.
• Route each request to whichever upstream account can serve the model you asked for.
• Meter tokens, deduct credit, and keep a per-call log you can reconcile against a bill.

What we do not do

• We do not train models, and we produce none of the output you receive. Every token comes from the vendor named in the model id.
• We do not claim to be, or to be endorsed by, any model vendor. Model names are used to describe what can be called, nothing more.
• We do not promise an SLA. There is no uptime guarantee and no service credit for downtime. If you need one, buy directly from the vendor.
• We do not sell data. There is no advertising network on this site and no broker on the other end of your account.

Why a gateway at all

A team that uses three vendors juggles three invoices, three keys, three rate limits and three failure modes. A gateway collapses that into one balance and one place to look when a call fails.

The trade is explicit: you take a third party in the path, and in exchange you get one bill and one integration. The section above says what that third party will not do for you.