# Relay: Self-Hosted LLM Gateway with Model Routing

A solution supporting OpenAI and Anthropic APIs with customizable routing policies.

- **Source:** [Hacker News: Show HN](https://news.ycombinator.com)
- **Published:** 2026-07-20
- **Topic:** [Data & AI](https://brifing.si/c/data-ai)
- **Original article:** <https://github.com/llmrelay/relay>
- **On Brifing:** https://brifing.si/s/3a6f0d81-3923-457f-85ff-b71e07e4994f

## What is Relay?

Relay is a self-hosted LLM gateway and model router that supports OpenAI and Anthropic APIs. It operates with a single Go binary and offers flexible routing policies. This allows developers to easily switch between different providers using their own API keys.

## Flexible Routing Policies

Relay provides a range of routing policies, from static aliases to learned smart layers. These policies can be customized according to user needs, balancing performance and cost.

## Evaluation Mechanism

Relay ensures routing decisions with an evaluation mechanism. This mechanism operates on models trained on user traffic, enhancing the reliability of smart routing.

## Privacy and Security

Relay collects no telemetry data and is entirely privacy-focused. This prevents user data from being shared with third parties, offering a secure solution.

## Easy Setup and Usage

Relay is a solution that can be set up in 60 seconds and requires zero configuration. Developers can quickly start using their own API keys and switch between various providers.

## Performance and Scalability

Relay stands out with low latency and high performance. While provider latency is the dominant factor, Relay's job is to minimize this latency and provide a seamless experience for users.

## Future and Development

Relay continues to develop its smart routing layers. These layers, trained on user traffic, aim to provide a more powerful and reliable solution in the future.

## Sources covering this story

- https://github.com/llmrelay/relay — 

---

_Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._