Get started with the AI Gateway

The AI Gateway is a managed entry point to large language models. It gives your organization an OpenAI-compatible API with centralized authentication, per-organization model access, usage metering, and spend controls — so your teams can build on approved models without managing provider keys themselves.

This guide walks you through the full path from zero to your first response.

For terminology used throughout the guides, see Key definitions.

The onboarding path

1. Request Organization — have an organization provisioned for your team.
2. Create a workspace — organize your team, applications, or workloads.
3. Add users to your organization — invite team members and assign organization roles.
4. Enable models — an org admin turns on the models you need.
5. Create credentials — mint an OAuth client for a workspace.
6. Get an access token — exchange the client for a bearer token.
7. Make your first API call — send a chat completion.
8. Monitor usage — track requests, tokens, and costs for your workspace.

What you'll need

• An organization on the AI Gateway (step 1).
• Admin access to a workspace in that org to create credentials (step 5).
• A client that can make HTTPS requests (curl, or any OpenAI-compatible SDK).

How access works

Every request is authenticated with an OAuth 2.0 bearer token and routed to your organization's dedicated path. The gateway checks that the token's client is authorized for your organization and that the requested model is enabled, then proxies the call to the upstream provider and records usage.

Note: URLs in this guide are for the non-production environment. Your platform administrator will give you the exact hostnames for your environment.