eyerouter — one gateway for every AI model
One OpenAI- and Anthropic-compatible address with automatic model fallback and exact billing per request.
What it is
eyerouter sits between an app and the AI models it uses. Developers call one address with a simple product name, and eyerouter picks the actual model behind it. It is in invite-only alpha.
The problem
Apps built on AI depend on providers that slow down, change prices or go offline. Switching provider usually means rewriting code, and the real cost of each request is hard to see.
How it works
Each product is backed by a ranked list of models; if one struggles, the request moves to the next healthy one and still completes. It speaks both the OpenAI and Anthropic formats, bills every request on actual usage in a ledger, and adds OCR for photos and PDFs plus document parsing for Word, Excel and PowerPoint files.
Why it matters
If customers depend on your AI answering, a fallback plan is not optional. Building it in from day one is cheaper than explaining an outage.
Scope
- Model routing & fallback
- OpenAI + Anthropic compatible
- Per-request billing
- OCR & document parsing
