multi-router.aidocs

API Reference

Models

GET /v1/models: the models available to the organization.

GET /v1/models lists the models the organization can use, in the OpenAI list format, so tools that read a model list work with Multi-Router. It needs the models:read scope. With an x-mr-router header naming a Claude Proxy, it lists that proxy's models in the Anthropic format instead. With a header naming an Adaptive Router, Codex Proxy, Gemini Proxy, fallback or weighted router, it returns 404.

The list

The list holds each active chat model in the catalog for the providers the organization has an enabled provider key for, and, with an enabled custom provider key, the installation's custom models. Each id is the model's catalog slug, which the model field accepts, and created is the catalog version's publish time:

Response 200json
{
  "object": "list",
  "data": [
    { "id": "anthropic/claude-opus-5", "object": "model", "created": 1790000000, "owned_by": "anthropic" },
    { "id": "anthropic/claude-sonnet-4.5", "object": "model", "created": 1790000000, "owned_by": "anthropic" },
    { "id": "openai/gpt-5.1", "object": "model", "created": 1790000000, "owned_by": "openai" }
  ]
}

A Claude Proxy's models

With x-mr-router naming a Claude Proxy, the response lists the models the proxy serves, in the Anthropic format: each entry has type, id as the provider's model id, display_name, created_at, max_input_tokens and max_tokens, with has_more, first_id and last_id. Claude Desktop and Claude Code use this list for their model pickers. An unknown router returns 404 the requested model or router was not found, and a disabled proxy 403 this Claude Proxy is disabled. A workload signed in with machine identity whose binding has several routers names one, or receives 400 router_required.

Proxy EndpointsCatalog
Was this helpful?