# Phase Router

> OpenAI-compatible model router: one API for every model, pay per call, a signed receipt for each call. Phase Router sends each request to the cheapest model that can do it well.

## Connect

- Base URL (OpenAI-compatible): https://ai.phase.xyz/v1
- Authentication: API key in the `Authorization: Bearer` header (keys start with `phk_`).
- Model `auto` lets the router pick per request; any model id from `GET /v1/models` calls that model.
- Remote MCP server: https://mcp.ai.phase.xyz/mcp (registry name `xyz.phase.ai/gateway`).

## Pay

- API key: prepaid balance, you pay the token price only.
- Wallet, no account (x402): pay each call in USDT on Phase L1; the price is quoted in a 402 before you sign, plus a per-call settlement fee. Details: https://ai.phase.xyz/docs#x402
- Your own provider key (Claude, GPT, Gemini, Grok): https://ai.phase.xyz/models?access=byok

## Receipts

- Every paid answer names the model that ran, its cost (`usage.cost`) and a task id; the signed receipt is at https://ai.phase.xyz/receipts/<task_id>
- Receipts roll into weekly epochs anchored on Phase L1: https://ai.phase.xyz/epochs

## Pages

- [Models and prices](https://ai.phase.xyz/models)
- [Docs](https://ai.phase.xyz/docs)
- [Chat](https://ai.phase.xyz/chat)
- [Sign up and API keys](https://ai.phase.xyz/dashboard)
- [Health](https://ai.phase.xyz/healthz)