Introduction

OriginRouter is a unified AI model gateway. A single API gives you access to dozens of mainstream models from OpenAI, Anthropic, Google, Zhipu, DeepSeek, and more — no need to integrate each provider separately.


What is OriginRouter

OriginRouter sits between your application and the underlying AI models as an intelligent routing layer. You call one standard HTTP API, and the system automatically routes each request to the best available channel based on your chosen model, subscription tier, and real-time network conditions.

We provide interfaces fully compatible with both the OpenAI API and the Anthropic API. That means:

  • If you're already using the OpenAI SDK or Anthropic SDK, just point base_url at our service — no changes to your business logic.
  • You can configure OriginRouter directly in tools like Claude Code, Cline, OpenCode, Chatbox, and SillyTavern.
  • A single API Key unlocks GPT, Claude, Gemini, GLM, DeepSeek, MiniMax, Kimi, and dozens more.

OriginRouter CLI is the open-source control layer running on your device. It preserves the native Claude Code and Codex experience while managing providers, model routes, remote sessions, approval policy, and protocol compatibility patches. OriginRouter App is the visual desktop and mobile workspace for chat, Provider Control, trusted devices, and remote agents. They work together over local direct connections or end-to-end encrypted trusted-device channels. See Products and entry points to choose a workflow.

With the OriginRouter One subscription, you also get access to OriginRouter Remote: connect to your terminal, Claude Code, or Codex environment from any device, anywhere.


Core Capabilities

Unified access to all models

One API Key covers every supported model. No separate registrations, top-ups, or key management per provider. Use GET /beta/v1/models to see the full list of currently available models at any time.

Full compatibility with mainstream ecosystems

  • OpenAI compatible: standard endpoints including /v1/chat/completions, /beta/v1/embeddings, and /beta/v1/images/generations — use the official OpenAI SDK directly.
  • Anthropic compatible: supports the /v1/messages endpoint, compatible with the Anthropic SDK and the Claude Code ecosystem.
  • Multi-platform: detailed configuration guides for Claude Code, OpenCode, cc-switch, Chatbox, SillyTavern, OpenClaw, and more.

Cost-effective coding plans

The OriginRouter One subscription is built for developers — top-tier coding models at a fraction of official pricing. For Claude Sonnet 4.6, the coding plan can reduce costs by about 85-95%, with monthly quota worth 10-15× the subscription fee.

Intelligent routing and high availability

The system automatically selects the optimal routing strategy based on your subscription tier, with multi-channel redundancy for stability. Direct, enterprise-grade distribution, and hybrid routing strategies balance performance against cost.

Advanced features

  • Multimodal adaptation: even when a target model doesn't support images or documents, the system automatically converts media content into text descriptions and injects them seamlessly into context.
  • Article Editing API: native document operations including session management, block editing, citation management, and style settings — embed document processing into your AI Agent.

Pick Your Endpoint

OriginRouter exposes three HTTP endpoint prefixes (/v1, /beta/v1, /coding/v1) and different SDK protocols need different base_url values. Pick your scenario below — Endpoint Configuration will show you the exact string to use:

Pay-as-you-go

Suited for product integrations: handle image and PDF inputs through one API, reuse uploaded files, and configure model fallback plus long-conversation memory.

Coding Plan

Subscription quota keeps cost lower and more predictable; mainstream coding models work out of the box with tools such as Claude Code and Codex CLI.

Pay-as-you-go · API access

For pay-as-you-go accounts. Use /beta/v1 for chat / messages / responses / images / embeddings / files / web.

Coding Plan · OriginRouter One

For OriginRouter One subscribers. Uses the dedicated coding-plan entry for chat / messages / responses / models.

Not sure which option applies? See Endpoint Configuration.


Two Ways to Use It

There are two typical ways to use OriginRouter. Understanding the difference will help you find the right docs quickly.

Option 1: Configure in a third-party tool or SDK

Set OriginRouter as the model provider in an AI tool or SDK you already use. Just swap the Base URL and API Key — no changes to your business code.

  • AI tools: Claude Code, Chatbox, OpenCode, SillyTavern, etc. — enter the OriginRouter address and key in the tool's settings.
  • SDKs: OpenAI Python/Node SDK, Anthropic Python/Node SDK, etc. — point base_url at OriginRouter when initializing the client.

Relevant docs: Third-party Platform Support lists compatible tools and SDKs. Each tool's detail page includes Base URL and API Key setup instructions.

Option 2: Call the API directly

Make HTTP requests directly to OriginRouter's API endpoints from your own code. You handle request construction, response parsing, and error handling yourself — ideal for developers embedding AI deeply into an application.

Relevant docs: The Model API section provides complete parameter references and code examples for every endpoint, including chat completions, embeddings, file upload, and image generation.


How to Use This Documentation

The docs are organized into five sections — jump to whichever fits your need:

0. Pick your Base URL — required reading for new users. Find the correct base_url string for your plan and SDK / tool (OpenAI SDK / Anthropic SDK / Claude Code / Codex CLI, etc.).

1. Quick Start (you are here) — what the platform does, which tools it supports, what models are available, and how to choose a plan.

2. Platform Support — step-by-step setup for each tool: Claude Code, cc-switch, OpenCode, Chatbox, SillyTavern, OpenClaw, and SDK integration guides.

3. Model List — complete list of all models available in the coding plan, with model IDs and routing strategies per subscription tier.

4. Model API — full reference for every API endpoint: chat completions, responses, messages, embeddings, files, images, web search, and error codes.

5. Article Editing API — WorkspaceLM native document operations: session management, block editing, citation management, and style settings.


Ready to get started?

Was this page helpful?