Skip to main content
AI gateways are currently in beta. Contact Opal support to enable them for your organization. If you self-host Opal, you need version 1.1252.0 or later.

​
Overview

LiteLLM works as an AI gateway for Opal when it routes Claude models to Anthropic. Opal sends its requests to LiteLLM’s Anthropic-compatible /v1/messages route, using the same model names it would send to Anthropic.

​
Prerequisites

  • A LiteLLM proxy that Opal can reach. On Opal cloud, it needs a public HTTPS URL.
  • An Anthropic API key for LiteLLM to forward requests with.

​
Configure LiteLLM

1

Route Claude models to Anthropic

In your LiteLLM config, map Claude model names to Anthropic. A wildcard keeps the names Opal sends unchanged:
Leave drop_params off, so LiteLLM doesn’t remove request settings Opal sends.
Don’t route Opal’s models to bedrock/ or vertex_ai/. Through Bedrock, LiteLLM passes the output schema as a hint instead of enforcing it, so the structured output check fails.
2

Create a key for Opal

Create a LiteLLM virtual key for Opal, so you can set its own budget and revoke it on its own. If you list allowed models on the key, include every model in the gateway requirements.

​
Connect it in Opal

Follow Connect your gateway with: LiteLLM’s /anthropic pass-through route also works. To use it, add /anthropic to the end of the URL.
Last modified on November 10, 2026