Migrate Claude Code to Volcano Engine and Unlock Free Tokens

On September 5, Anthropic announced it will stop providing Claude services to enterprises and subsidiaries in China that are controlled by capital holdings over 50%.

In response, Volcano Engine quickly launched a migration solution, allowing developers to call Claude Code models on Volcano Engine or combine TRAE with Volcano Ark to achieve equivalent capabilities.

Volcano Engine also offers multiple incentives: 3 million free Tokens plus a 50% discount on up to 1 billion Tokens.

New users receive 3 million Token vouchers and a limited 50% discount per model.

Joining the data reward collaboration plan grants individual users an extra 500,000 Tokens per model per day and enterprise users 5 million Tokens per model per day.

Solution 1: Claude Code with Volcano Ark models

Install Claude Code and the Claude Code Router, then create a configuration file (~/.claude-code-router/config.json) with the appropriate provider and router settings.

{
  "LOG": false,
  "OPENAI_API_KEY": "",
  "OPENAI_BASE_URL": "",
  "OPENAI_MODEL": "",
  "Providers": [
    {
      "name": "bytedance-ark",
      "api_base_url": "https://ark-cn-beijing.bytedance.net/api/v3/chat/completions",
      "api_key": "{Your API Key}",
      "models": [
        "{Your Model Name}"
      ]
    }
  ],
  "Router": {
    "default": "ark,{Your Model Name}",
    "think": "ark,{Your Model Name}",
    "background": "ark,{Your Model Name}",
    "longContext": "ark,{Your Model Name}"
  }
}

After editing the config file, restart the router with ccr restart, then launch Claude Code using ccr code and confirm with “yes”.

Solution 2: TRAE + Volcano Ark

In TRAE (The Real AI Engineer), developers can copy or import the Claude Code directory into the TRAE environment and continue using Volcano Ark models. TRAE fully supports MCP protocols, providing a seamless migration while retaining existing workflows and offering localized services.

TRAE is an AI IDE tailored for Chinese developers, with over one million monthly active users and top rankings on the Swebench leaderboard. It supports multi‑language, multi‑framework, and edge scenarios, offering Chat, Builder, and Cue modes for code Q&A, task decomposition, and code completion.

Chat mode: code Q&A based on internal repositories and APIs.

Builder mode: generate task breakdowns and code suggestions from natural language.

Cue mode: real‑time code completion with high accuracy.

TRAE also leverages the multimodal capabilities of Doubao Model 1.6 and integrates with Volcano Ark’s MCP Hub for end‑to‑end AI application development.

Volcano Ark continues to add high‑quality models such as DeepSeek V3.1, Doubao Model 1.6, and Kimi K2, offering low latency (20‑40 ms per token), high concurrency, and a full AI development toolchain.

Developers can experience these services in the Volcano Engine experience center.