Step-by-Step Guide: Integrate a Chinese Large Model into Claude Code for a Low-Cost AI Terminal

Step 1: Install Node.js

Claude Code runs on Node.js, so download the Windows installer from the official Node.js website (https://nodejs.org/zh-cn/download) and follow the wizard to completion. After installation, open the Windows Command Prompt, run node -v, and verify that a version string (e.g., v24.11.0) is displayed.

Step 2: Install Git

Git is required for pulling projects and version management. Visit the official Git website (https://git-scm.com/install/windows), click the "Git for Windows/x64 Setup" button, download the installer, and run it to finish.

Step 3: Install Claude Code

Open Command Prompt again and execute the following npm command: npm install -g @anthropic-ai/claude-code The installation finishes with a Claude Code prompt. Run claude, choose a theme with the arrow keys, and you will see the Claude Code interface, confirming a successful install.

Step 4: Install cc‑switch and configure a domestic model

The official Claude Code requires overseas access and a paid subscription (US$20‑$200 per month). To avoid this, replace its backend with a Chinese large model. Download the open‑source cc-switch tool from its GitHub releases (e.g., CC‑Switch‑v3.6.2‑Windows‑Portable.zip), unzip the package, and launch the executable.

In the cc‑switch window, click “Add Provider”, paste the API key of the chosen domestic model (options include Kimi, Tongyi, Zhipu GLM, DeepSeek), and enable the model.

Step 5: Activate the custom model in Claude Code

Restart Claude Code, run claude again, and when prompted “Use custom API KEY?”, answer **Yes**. Confirm the model status is enabled in the cc‑switch interface.

Step 6: Test with small projects

Project 1 – “Space Invaders” game : Ask Claude Code to generate a simple Space Invaders game. The tool produces a runnable game in under a minute, and the resulting UI is displayed.

Project 2 – Real‑time quiz web app : Request a multi‑user, real‑time answering platform with leaderboard and teacher control. After about twenty minutes Claude Code returns a complete web application that supports 100 concurrent users.

Conclusion : By swapping the backend to Zhipu GLM (or another domestic model) the Claude Code terminal runs smoothly, responds quickly, and avoids VPN requirements and high subscription fees, offering a cost‑effective AI coding assistant.