CCG

Claude Code Router Go

A lightweight Go implementation of Claude Code Router

Smart Routing for AI Models

English | 中文文档

---

About

CCG is a Go-based implementation of CCR. Due to WSL frequently crashing from high memory usage, this project was created to provide a lightweight alternative.

Key Benefits

• Low Memory: ~30MB vs ~150MB (Node.js version)
• Fast Startup: ~0.1s vs ~2s (Node.js version)
• No Dependencies: Single binary, no runtime required

---

The Story Behind CCG

This project is a product of Vibe Coding.

I don't know how to write Go code. Not a single line was written by hand - the entire codebase was generated through conversations with AI. I just described what I needed for my WSL environment and let AI handle the implementation. (Yes, this sentence was also written by AI.)

---

Installation

# One-line install
curl -fsSL https://raw.githubusercontent.com/visvioce/ccg/master/install.sh | bash

# Build from source
git clone https://github.com/visvioce/ccg.git
cd ccg
make install

---

Quick Start

1. Configure environment variables:

   export OPENAI_API_KEY="your-key"
   export DEEPSEEK_API_KEY="your-key"
   # Add other provider keys as needed

2. Start the server:

   ccg start

3. Open Web UI:

   ccg ui

---

Commands

Server Management

Command	Description
ccg start	Start server
ccg stop	Stop server
ccg restart	Restart server
ccg status	Show server status
ccg ui	Open Web UI in browser

Information

Command	Description
ccg model	List available models
ccg env	Show environment variables
ccg activate	Output env vars for shell
ccg version	Show version

Execution

Command	Description
ccg code <prompt>	Execute AI command

Presets

Command	Description
ccg preset list	List presets
ccg preset info <name>	Show preset info
ccg preset install <source>	Install preset
ccg preset export <name>	Export preset
ccg preset delete <name>	Delete preset

Marketplace

Command	Description
ccg install <name>	Install preset from marketplace

---

Configuration

File: ~/.ccg/config.json

{
  "HOST": "127.0.0.1",
  "PORT": "3456",
  "providers": [
    {
      "name": "openai",
      "api_base_url": "https://api.openai.com/v1/chat/completions",
      "api_key": "${OPENAI_API_KEY}",
      "models": ["gpt-4o", "gpt-4o-mini"],
      "transformer": { "use": [] }
    }
  ],
  "Router": {
    "default": "openai,gpt-4o",
    "background": "openai,gpt-4o-mini",
    "think": "openai,gpt-4o",
    "longContext": "openai,gpt-4o",
    "longContextThreshold": 60000,
    "webSearch": "openai,gpt-4o"
  }
}

CCG works with any OpenAI-compatible API. Configure providers in the providers array.

---

Routing

Routes requests based on scenario:

Scenario	Description
default	General requests
background	Low-priority tasks
think	Complex reasoning
longContext	Long context (>threshold tokens)
webSearch	Web search

---

API

POST http://localhost:3456/v1/chat/completions
GET  http://localhost:3456/v1/models

---

License

MIT License