How to Use Claude Code in China: Three Paths

Word count: ~2100 words

Estimated reading time: ~10 minutes

Last updated: August 1, 2025

Is this article for you?

✅ You’re a developer in China eager to use one of the most powerful coding AIs — Claude Code.

✅ You’ve tried registering or using it yourself but got discouraged by complex network environments and strict account policies.

✅ You want to find a stable, reliable solution suited to your needs rather than repeatedly troubleshooting.

✅ You’re seeking a clear decision guide rather than scattered method listings.

Chapter Contents

Preface: Facing the Core Contradiction
Path 1: The Stable Choice (Quality Proxy Solution)
Path 2: The Hardcore Challenge (Self-Managed Network Environment)
Path 3: The Flexible Workaround (Open-Source Model Alternatives)
Summary: How to Make Your Optimal Decision

Preface: Facing the Core Contradiction

Hey, I’m Mr. Guo. Claude Code’s power needs no further elaboration. Its deep code understanding and complex engineering mastery make it a core tool in many top developers’ AI toolboxes. But a very practical problem stands before us: for the vast majority of Chinese developers, using it smoothly is extremely difficult.

The core contradiction is that a top-tier productivity tool has its usage barrier raised to unreasonable heights by network environment and account policies. As an efficiency maximalist, I believe developers’ precious time and energy should be invested in solving valuable business problems, not consumed in repeated “battles” with network and account issues.

Therefore, this article’s purpose is to clear the fog and provide three clear, explicit paths I’ve analyzed and verified. They correspond to different user profiles and risk preferences. Please make the strategic choice best suited to your actual situation.

Path 1: The Stable Choice (Quality Proxy Solution) 🚀

This is my most recommended path — the one most aligned with the principle of “leave professional matters to professionals.”

Target Users: The vast majority of developers, especially those wanting “out-of-box” solutions, unwilling to waste time on network issues, and willing to pay reasonable fees for stable service.
Core Logic: Through a reliable third-party service provider who handles all the complex network and account issues for you. You just need their provided credentials and simple configuration to directly use native Claude Code, enjoying the identical top-tier experience.
Pros: Extremely stable — no worrying about IP pollution or account bans; Simple to use — typically with detailed Chinese tutorials; Zero maintenance cost — configure once, use long-term.
Cons: Requires paying service fees.

Mr. Guo’s Practical Solution:

Recommended Service: code.yoretea.com I’ve been using this solution long-term with excellent experience. It perfectly solves the core stability pain point, and their website’s tutorial documentation is excellently done — extremely beginner-friendly.

🎁 Mr. Guo’s Exclusive: 30% off code GUOSHU, applies to all products.

Path 2: The Hardcore Challenge (Self-Managed Network Environment) 🛠️

This path suits “hardcore players” with full confidence in their technical abilities and network resources.

Target Users: Technical experts with stable, clean, dedicated overseas IP resources who are familiar with network configuration.
Core Logic: Solve network issues yourself, simulating being a genuine overseas user to directly access Anthropic’s official website for registration and use.
Pros: Direct interaction with official services, no middleman. If you already have network resources, potentially lower cost.
Cons: Extremely high ban risk — Anthropic’s risk control system is very strict, with near-zero tolerance for shared or dirty IPs; High maintenance cost — you need to continuously ensure your network environment stays stable and reliable, which itself is time and energy consuming.

Steps and Warning:

Ensure you have a high-quality dedicated IP in a supported region (US, UK, etc.).
Enable global network proxy, visit the official site to register and subscribe.
In your local terminal, you also need to configure the proxy environment before installing and logging into Claude Code CLI.

Serious Warning: This path is unsuitable for the vast majority of users — its “fragility” far exceeds your imagination. Any single point of failure could cause your account and investment to go up in smoke.

Path 3: The Flexible Workaround (Open-Source Model Alternatives) 🧩

This path’s core idea is “decoupling”: we appreciate Claude Code’s excellent CLI workflow, but we can swap the underlying large model for other equally excellent models more easily accessible in China.

Target Users: Users pursuing extreme cost-effectiveness, who enjoy tinkering, or have special needs for domestic LLMs’ (like Kimi) long-context capabilities.
Core Logic: Using scripts provided by the open-source community, “mod” Claude Code’s client so its backend requests point to other models’ API endpoints instead of Anthropic’s servers.
Pros: Completely unrestricted by network; Extremely low cost — can utilize free quotas from models like Kimi; Excellent workflow experience — retains most of Claude Code CLI’s interaction experience.
Cons: Not authentic Claude — code generation quality and style depend on the model you swap in; relies on community projects that may break with official updates.

Steps (Using Kimi as Example):

Install Node.js and original Claude Code CLI (npm install -g @anthropic-ai/claude-code).
Go to Kimi’s open platform to apply for an API Key and add credit.
Execute the open-source project kimi-cc’s installation script — it auto-completes configuration.
Restart terminal, and you can start using the claude command — now powered by Kimi model behind the scenes.

🏁 Summary: How to Make Your Optimal Decision

Which path to choose depends on your comprehensive weighing of stability, cost, risk, and willingness to tinker. Here’s a simple decision checklist:

If you pursue “stability above all” and want to maximize time value, Path 1 is your best choice.
If you’re a network expert unafraid of risk and enjoy controlling everything, challenge Path 2.
If you’re an open-source enthusiast, cost-sensitive, and enjoy trying different model combinations, Path 3 opens a new door.

In the AI era, choosing the right tool combination and workflow is as important as learning a new programming language. Hope this article helps you make the wisest choice, letting you truly convert Claude Code’s powerful capabilities into hands-on productivity.

Found Mr. Guo’s analysis insightful? Drop a 👍 and share with more friends who need it!

Follow my channel to explore AI, going global, and digital marketing’s infinite possibilities together.

🌌 The best tool is one that lets you forget the tool itself and focus on creating.