SwatowAPI

English

简体中文

English

简体中文

Appearance

Introduction ​

SwatowAPI (汕拓智算) is a multi-model AI gateway. One endpoint, 50+ models, compatible with the OpenAI, Anthropic, and Gemini protocols at the same time. Keep your existing SDKs and workflow; just point the Base URL and key at SwatowAPI.

Three core ideas ​

  • Multi-model routing: one set of compatible endpoints fronts 50+ models across GPT, Claude, Gemini, DeepSeek, Qwen, GLM, Kimi, and more. No need to wire up a separate SDK per vendor.
  • Auto-route by cost and availability, with fallback on failure: the gateway routes by cost and availability, and falls back to a working channel when an upstream errors or times out, transparently to the caller.
  • Transparent ledger, traceable per call: every call carries a trace id, and input/output tokens are metered against official pricing with no markup. Failed and timed-out requests are not billed, top-ups and spend reconcile in one ledger, and channel status is public.

Key concept: keys, groups, and protocol surface ​

This is the one thing to understand before you start.

Every API key belongs to a group, and a group has a platform: OpenAI, Anthropic, or Google. Which protocol surface you use is determined by the group platform of the key you call with.

  • The same endpoint auto-routes based on your key's group platform.
  • A key whose group platform is OpenAI uses the OpenAI-compatible protocol (for example /v1/chat/completions, /v1/images/generations).
  • A key whose group platform is Anthropic uses the Anthropic-compatible protocol (for example /v1/messages).
  • A key whose group platform is Google uses the Gemini-compatible protocol (/v1beta/...).

So calling an OpenAI-only endpoint (such as image generation) with a non-OpenAI group key returns 404, and vice versa. Confirm your key's group platform first, then pick the matching protocol surface.

TIP

Not sure of a model's exact model ID? Call GET /v1/models and use what the server returns.

Who it's for ​

  • Developers who want one key and one endpoint across many vendors' models, without maintaining multiple SDKs.
  • Applications that want auto-routing by cost and availability with fallback on failure for resilience.
  • Teams that need a per-call traceable, reconcilable usage ledger.
  • Projects already on the OpenAI, Anthropic, or Gemini SDK that just want to swap the Base URL for a smooth migration.

Next steps ​