Providers

Base URL recommendations

For OpenAI-compatible relay services, the safest Base URL is usually the provider endpoint with /v1 included.

Examples:

https://example.com/v1
https://api.example.com/v1

Clash for AI has compatibility handling for model list requests, but using the documented /v1 Base URL is still the clearest setup for end users.

Model list behavior

Different relay services expose model lists differently.

Common issues:

1. The provider may not expose a model list endpoint at all.
2. The provider may only respond at /v1/models, not /models.
3. The returned JSON may differ from standard OpenAI-compatible formats.

Because of this, model list fetching should be treated as a compatibility feature, not as a guaranteed capability of every relay provider.

Authentication notes

Clash for AI currently focuses on OpenAI-compatible and Anthropic-compatible upstream integrations.

That means:

1. OpenAI-style Authorization: Bearer ... flows are supported.
2. Anthropic-style x-api-key flows are supported.
3. Gemini native protocol is not currently supported as a first-class upstream protocol.