---
title: "PipeLLM Relay Auto Router"
route_path: "/relay"
canonical_url: "https://www.pipellm.ai/relay"
markdown_path: "/llms/relay.md"
markdown_url: "https://www.pipellm.ai/llms/relay.md"
content_type: "product-page"
description: "Product page for PipeLLM Relay Auto Router, including protocol translation, policy routing, provider failover, model discounts, and SDK compatibility."
generated_at: "2026-08-22T04:42:12.577Z"
---
Canonical page: https://www.pipellm.ai/relay
Markdown mirror: https://www.pipellm.ai/llms/relay.md
Content type: product-page
Generated at: 2026-08-22T04:42:12.577Z
# PipeLLM Relay Auto Router

Relay is PipeLLM's governed Auto Router. It sits between production applications and AI model providers so teams can keep familiar SDKs while automatically selecting approved routes by policy, provider health, region, availability, and cost.

## What Relay Does

- Provides OpenAI-, Anthropic-, and Gemini-compatible request paths.
- Authenticates each request and applies team, key, budget, model, region, and tool policies.
- Selects an approved model and provider route based on policy, health, latency, region, availability, and cost.
- Streams a normalized response back through the client's existing SDK interface.
- Connects routing decisions to PipeLLM Lens traces for observability and audit.

## Production Benefits

- Change or add model providers without rewriting each application integration.
- Keep experimental or unapproved models out of production paths.
- Fail over between approved upstream providers when availability changes.
- Manage provider credentials and access outside application source code.

## Get Started

- Open Relay: https://console.pipellm.ai
- Browse models: https://www.pipellm.ai/models
- Read the API documentation: https://docs.pipellm.ai/api-reference/introduction