To set up Claude Code to run through an OpenAI-compatible API server using a tool

like Claude Code Router (CCR), you need to configure a local proxy that translates

Claude’s native Anthropic requests into standard OpenAI API payloads

配置CCR

创建配置文件~/.claude-code-router/config.json

{
  "LOG": true,
  "PORT": 3456,
  "LOG_LEVEL": "info",
  "Providers": [{
    "name": "gemini",
    "api_base_url": "https://generativelanguage.googleapis.com/v1beta/openai/",
    "api_key": "${GEMINI_API_KEY}",
    "models": ["gemini-2.5-flash", "gemini-2.5-pro"],
    "transformer": { "use": ["gemini"] }
  }],
  "Router": {
    "default": "gemini,gemini-2.5-flash",
    "background": "gemini,gemini-2.5-flash",
    "image": "gemini,gemini-2.5-flash",
    "think": "gemini,gemini-2.5-flash"
  }
}

启动服务

ccr start
ccr status

测试

import os

from openai import OpenAI


client = OpenAI(
    api_key=os.getenv("GEMINI_API_KEY"),          # 从 https://aistudio.google.com/apikey 获取
    base_url="https://generativelanguage.googleapis.com/v1beta/openai/"
)

response = client.chat.completions.create(
    # model="gemini-2.5-pro-exp-03-25",      # 常用模型推荐
    model="gemini-3-flash-preview",            # 速度快、免费额度高
    # model="gemini-1.5-pro",              # 老模型
    messages=[
        {"role": "system", "content": "你是一个有帮助的AI助手。"},
        {"role": "user", "content": "介绍一下你自己"}
    ],
    temperature=0.7,
    max_tokens=2048
)

print(response.choices[0].message.content)