Real-time paid streaming access to every major AI model, powered by an x402-API gateway and deployed on Cloudflare Workers.  

Each inference request is billed in stablecoins on the Radius Network using EIP-2612 permit signatures, settling on-chain in ~500ms. A lightweight local proxy (rad-router-proxy) intercepts standard OpenAI-compatible calls from coding agents like Claude Code, Zed, or Kilo Code, auto-signs payment permits, and forwards them to RadRouter, which verifies and settles the payment before streaming the response back from OpenRouter.