API compatibility and selection

Connect through OpenAI-compatible endpoints and native client APIs

XiuRouter supports Responses, Chat Completions, Anthropic Messages, and Gemini GenerateContent. Choose the API format your client uses, then follow the matching setup guide.

When to use an OpenAI-compatible endpoint

Use it when your application lets you set the Base URL and API key.

  • Your OpenAI SDK or compatible client lets you set its Base URL and API key.
  • You want to choose models and service tiers from the current XiuRouter catalog.
  • You can test one request and check it in Usage before production use.

Choose the API format your client uses

XiuRouter supports several request formats. The Base URL and path must match the client or SDK.

responses

OpenAI Responses

Base URL
https://router-api.xiu.ai/v1
Request path
/responses

Use with Codex, agents that send OpenAI Responses requests, and new applications built on Responses.

The current endpoint is stateless. Do not assume provider-hosted stored conversations, background mode, web search, or file search are available.

messages

Anthropic Messages

Base URL
https://router-api.xiu.ai
Request path
/v1/messages

Use with Claude Code, Anthropic SDKs, and Claude-native clients that send Messages requests.

Messages compatibility does not imply every Anthropic platform feature. Hosted server tools, Message Batches, and Files are outside the current scope.

chat

OpenAI Chat Completions

Base URL
https://router-api.xiu.ai/v1
Request path
/chat/completions

Use with clients limited to Chat Completions, existing OpenAI-compatible tools, and supported custom-model integrations.

Tool calling, streaming, and client-specific agent behavior still depend on the client and selected model; the compatible route alone cannot guarantee them.

gemini

Gemini GenerateContent

Base URL
https://router-api.xiu.ai
Request path
/v1beta/models/YOUR_MODEL_ID:generateContent

Use with Gemini SDKs and clients that require the native Gemini GenerateContent request format.

The model ID is part of the request path, and the Base URL differs from OpenAI SDKs. Do not add /v1 unless the matching guide shows it.

Choose in three steps

Start with the API format used by the client, then confirm the model and test one request.

  1. 1

    Identify the client protocol

    Check whether the client sends Responses, Chat Completions, Anthropic Messages, or Gemini GenerateContent.

  2. 2

    Confirm the model and service tier

    Use the current Models and pricing page for the exact model ID, declared endpoint types, tier, and unit rates.

  3. 3

    Verify one real request

    Create a scoped key, send a small request, and confirm the model, tier, tokens, status, and recorded cost in Usage.

Next steps