Add API Route as a first-class OpenAI-compatible gateway

Summary

Add API Route as a first-class OpenAI-compatible gateway in OpenClaude.

API Route is an OpenAI-compatible multi-model AI API gateway that provides access to GPT, Claude, Gemini, Kimi, Qwen, and other model families through one API key.

Website: https://www.api-route.com/

Documentation: https://www.api-route.com/docs/overview

Quickstart: https://www.api-route.com/docs/quickstart

Problem

API Route can already work with OpenClaude through the generic OpenAI-compatible route, but users currently need to configure the endpoint and credentials manually.

A first-class gateway preset would improve discoverability and reduce setup friction by giving API Route its own entry in /provider, dedicated credential handling, and model discovery.

Proposed Direction

Add an official API Route gateway descriptor and preset.

Suggested scope:

• Gateway ID: api-route
• Gateway label: API Route
• Transport: openai-compatible
• API key env: API_ROUTE_API_KEY
• Authentication: Bearer token
• Use API Route's OpenAI-compatible Base URL from its API Access configuration
• Reuse OpenClaude's existing OpenAI-compatible chat, streaming, and tool-calling transport
• Discover available models through API Route's OpenAI-compatible model catalog
• Expose API Route as a preset in /provider
• Support saved provider profiles and environment-based configuration
• Add a short setup guide and gateway documentation

No new transport or inference protocol should be required. The integration should follow OpenClaude's existing gateway descriptor architecture and reuse the existing OpenAI-compatible transport.

Why this is useful

API Route provides a single API key and OpenAI-compatible interface for multiple model families, including GPT, Claude, Gemini, Kimi, and Qwen.

This makes it useful for OpenClaude users who want to switch between multiple models without maintaining separate upstream accounts and endpoint configurations.

Alternatives

Users can continue configuring API Route manually as a custom OpenAI-compatible route, but this requires additional setup and makes the gateway less discoverable than existing first-class gateway integrations.

If this integration fits the project direction, I'm happy to prepare and test a focused PR following OpenClaude's existing gateway descriptor architecture.