Skip to content

API Overview

WRouter speaks the industry's mainstream LLM protocols in parallel — pick whichever you want.

Base URL

https://wrouter.ai

All endpoints are HTTPS only.

Protocol compatibility

ProtocolBase pathNotes
OpenAI-style/v1/*The industry's most common protocol format — official and community SDKs work as drop-in
Anthropic-style/v1/messagesMessages protocol format, for SDKs and clients that speak it
Gemini-style/v1beta/models/{model}:generateContentGenerative AI protocol format

Point your SDK's base_url at WRouter, leave the model ID and request body as-is, and WRouter routes to the right upstream.

Endpoints

Chat

Retrieval

Images

Audio

  • Speech (TTS)POST /v1/audio/speech
  • Transcriptions (STT)POST /v1/audio/transcriptions
  • Translations — POST /v1/audio/translations (any language → English)

Video

  • Video generation — POST /v1/videos and vendor-specific Kling / Pixverse / Seedance / Veo3 / Wanx / Jimeng entries
  • Task lookup — GET /v1/videos/{task_id}, GET /v1/videos/{task_id}/content

Safety

Realtime

  • Realtime — GET /v1/realtime (WebSocket upgrade)

Metadata

  • ModelsGET /v1/models, GET /v1beta/models

Conventions

  • Auth — All requests need Authorization: Bearer sk-.... See Authentication.
  • Content type — JSON requests are application/json; charset=utf-8 except file uploads.
  • Errors — Standard RESTful codes. See Errors.
  • Streaming — Every chat endpoint accepts stream: true (Server-Sent Events).
  • Rate limits — Per-token by default; configurable in the console.
  • Timeouts — Set client timeout ≥ 600 s; long generations may exceed 5 minutes.