---
title: "The Best Router for LLMs"
description: "An LLM router built on edge gateway architecture: one unified API with native OpenAI, Anthropic and Gemini protocols, every request routed to the best upstream. Pay as you go, no deposit fee."
canonical: https://pirouter.ai/
last-updated: 2026-08-30
---

# The Best Router for LLMs

> An LLM router built on edge gateway architecture: one unified API with native OpenAI, Anthropic and Gemini protocols, every request routed to the best upstream. Pay as you go, no deposit fee.

Smart router, better pricing, better uptime.

## Why PiRouter

### Routes across every provider

Each model runs on multiple upstreams. We check their health continuously and send your call to the fastest one — and away from the one that just broke.

### Works with your native SDK

OpenAI, Anthropic, and Gemini protocols all terminate here. Change the base URL, keep everything else.

### Team controls, standard

Shared org wallet, per-key rate limits, per-key provider allowlists. All standard. Email us when you need SSO or a custom SLA.

### Edge routing, direct to the provider

Your call enters at the nearest edge and goes straight to the provider from there. No detour through a central region.

## Up and running in minutes

### Sign up

Create an account in the Console. Add your team later with a shared org wallet.

### Buy credits

Top up your wallet once and spend it across every model, metered per token.

### Get your API key

Create a key and start calling with the SDK you already use — OpenAI, Anthropic, or Gemini.

## Questions, answered

### How do I integrate? How much code changes?

Keep your existing OpenAI / Anthropic / Gemini SDK. Point the base URL at api.pirouter.ai, swap in your PiRouter API key, and leave the rest of your code alone.

### Which models are supported?

Only models that pass our evaluation make the list. See the model catalog for the full lineup — it updates as evaluations evolve.

### Are there rate limits? Can they be raised?

Rate limits are configured per API key, with default rpm quotas per plan. Need more? Email us with your use case.

### How do I debug a failed request?

Every response carries a request id. Send the request id, timestamp, and symptoms to support — we trace it through the call chain.

## Explore

- https://pirouter.ai/models
- https://pirouter.ai/pricing
- https://pirouter.ai/sdk
- https://pirouter.ai/data
- https://pirouter.ai/llms.txt

---

HTML version: https://pirouter.ai/ · Machine-readable index: https://pirouter.ai/llms.txt