Hydedocs

Introduction

What Hyde Inference is and how it fits your setup.

Hyde Inference is an API gateway that serves large language models over the OpenAI API format.

What it is

You send requests to a single endpoint (https://inference.hyde.tech/v1) in the same shape the OpenAI API uses. The gateway routes each request to the model you asked for and returns the response in the OpenAI response format. Any tool or SDK that already speaks OpenAI works without code changes.

Every request is authenticated with a Hyde API key that you create in the dashboard. The dashboard shows usage, token counts, and latency for your keys under Overview and Usage.

How you use it

There are two common setups.

Your own code: call the API directly with the OpenAI SDK (Python, TypeScript, etc.) or plain HTTP. See the Quickstart.

Coding agents: point Claude Code, Cursor, OpenCode, or GitHub Copilot at the gateway. Each tool has a short guide with the exact config:

Models

The gateway currently serves hyde/glm-5.3-flash, hyde/kimi-k3, and hyde/kimi-k3.1-pro for chat, and hyde/qwen3-embedding-8b and hyde/nemotron-3.5-embed-8b for embeddings (/v1/embeddings). Context windows, output limits, and what each model is best at are on the Models page.

Start here

The Quickstart walks you through creating a key and sending your first completion.

Was this page helpful?