Unlock Multi‑Model AI Routing with Claude Code Router: A Complete Setup Guide

Why Claude Code Router?

Claude Code requires an Anthropic account and can become expensive due to API calls and token usage. Cheap models often fail on complex tasks, and many promoters sell hype rather than practical tools. Claude Code Router removes the account barrier and lets you route Claude Code requests to providers such as DeepSeek, OpenRouter, and Kimi, achieving true model freedom.

Key Advantages

No need for an Anthropic official account.

Supports major LLM providers.

Flexible model‑routing strategies (default, background, think, long‑context, web‑search).

Real‑time status line showing directory, Git branch, model type, and token consumption.

Custom transformer plugins for extensibility.

Quick Deployment Guide

Environment Preparation

Install nvm (Node version manager) to handle Node.js versions:

curl -o- https://gitee.com/mirrors/nvm/raw/v0.39.7/install.sh | bash
# If curl is missing:
# yum install -y curl   # CentOS
# apt install -y curl   # Ubuntu

Refresh the environment and verify:

source ~/.bashrc   # or source ~/.zshrc for zsh
nvm --version

Install Node.js (e.g., version 22):

nvm install node          # latest
nvm install 22            # specific version
nvm alias default 22       # set default

Check versions:

node -v   # e.g., v22.21.3
npm -v

Install Claude Code and Router

Install the original Claude Code CLI: npm install -g @anthropic-ai/claude-code Install Claude Code Router: npm install -g @musistudio/claude-code-router The installation requires no manual dependency handling.

Configuration

Create or edit ~/.claude-code-router/config.json. A minimal example:

{
  "LOG": false,
  "HOST": "0.0.0.0",
  "API_TIMEOUT_MS": 600000,
  "Providers": [
    {
      "name": "deepseek",
      "api_base_url": "https://api.deepseek.com/chat/completions",
      "api_key": "sk-xxx",
      "models": ["deepseek-chat", "deepseek-reasoner"],
      "transformer": {"use": ["deepseek"], "deepseek-chat": {"use": ["tooluse"]}}
    },
    {
      "name": "gemini",
      "api_base_url": "https://generativelanguage.googleapis.com/v1beta/models/",
      "api_key": "xx",
      "models": ["gemini-2.5-flash", "gemini-2.5-pro"],
      "transformer": {"use": ["gemini"]}
    }
  ],
  "Router": {
    "default": "deepseek,deepseek-chat",
    "background": "deepseek,deepseek-chat",
    "think": "deepseek,deepseek-reasoner",
    "longContext": "gemini,gemini-2.5-pro",
    "longContextThreshold": 60000,
    "webSearch": "gemini,gemini-2.5-flash"
  }
}

Routing rules explained: default: model for ordinary tasks. background: model for background jobs (cheaper local models possible). think: model for inference‑intensive tasks. longContext: model supporting >60K tokens. longContextThreshold: token threshold to trigger long‑context mode (default 60000). webSearch: model for web‑search tasks (add :online suffix for OpenRouter).

Running and Managing

Use the CLI to run Claude Code through the router: ccr code After changing the configuration, restart the service: ccr restart Open the UI for visual management (replace xxxx with your host IP):

http://xxxx:3456/ui

Conclusion

Claude Code offers a great user experience but incurs high subscription and network costs. Claude Code Router provides a low‑cost, flexible alternative that retains Claude Code’s workflow while allowing seamless integration with any LLM provider, improving freedom, controllability, and overall development efficiency.