How it works

Roteo sits between your app and the AI models. You call one endpoint with one key; Roteo routes the request to a model, returns the real output, and settles the cost onchain from your wallet. No subscriptions, no prepaid credits, nothing kept.

The request path

1. You send a request. One base URL, one sk_ key, a standard Messages or image request.
2. Roteo routes it. The Gateway is a routing and aggregation layer over multiple model providers. It picks an efficient path and reroutes automatically if a provider stumbles.
3. The model runs. The output is generated by the model and returned to you exactly as generated, byte for byte.
4. You pay per call. The cost settles onchain from your own wallet over x402, capped by your balance. You only hold stablecoins; Roteo covers the network gas.

Where the inference comes from

Every call is served by the same frontier models you would reach going direct. Rather than pay per-call list price, Roteo sources inference capacity in bulk across providers and routes each request down the most efficient path, then passes the wholesale rate through to you. That, together with onchain settlement (no card processors or subscription margins), is why the rates sit below going direct. See the FAQ for more.

What Roteo never does

• Store your content. Prompts and responses are never kept, and the web chat stays in your browser. See Privacy.
• Hold your funds. Your wallet is self-custodied and only you control it. See Pay-per-call billing.
• Change the output. What comes back is the model's genuine output, with nothing rewritten or injected.

Ready to try it? Head to the Quickstart.