Claude Code Getting Started Guide: Full Installation, Configuration, and Hands-On Walkthrough

What Is Claude Code: An AI Programming Assistant That Actually Does the Work

Claude Code has been gaining serious traction in developer communities lately. It's an AI coding assistant, but fundamentally different from familiar web-based chat tools like ChatGPT, DeepSeek, or others — it doesn't just chat, it can take direct action on your computer.

Just tell it what you need in plain language, and it understands your project structure and executes tasks automatically. This "autonomous execution" capability is the core differentiator setting it apart from traditional AI tools.

Five Dimensions That Set Claude Code Apart from Ordinary AI

There are clear differences between standard AI web chat and Claude Code across five key dimensions:

• Interaction style: Ordinary AI requires you to repeatedly copy code, paste it into a chat box, then paste it back into your editor — a tedious back-and-forth. Claude Code operates directly inside your project, eliminating that friction.
• Context awareness: Ordinary AI only knows what you explicitly tell it; Claude Code can automatically read your entire codebase and search files on its own.
• Execution capability: Ordinary AI gives suggestions or code snippets; Claude Code can directly create files, modify code, run commands, and execute tests.
• Memory mechanism: Ordinary AI forgets everything when you close it; Claude Code remembers project rules and your coding preferences through configuration files.
• Tool integration: Ordinary AI struggles to connect with external tools; Claude Code can interface directly with browsers, databases, GitHub, and other services via MCP.

A useful analogy: traditional AI is like a remote consultant you call for advice. Claude Code is more like an assistant sitting right next to you — one who can flip through your folders and edit your code themselves.

Claude Code vs Cursor: Is It Worth Switching?

Many developers face a practical question: if you're already using Cursor, is it worth switching to Claude Code?

From real-world usage, both tools handle everyday development tasks well — use whichever feels more natural. The main reason some developers have recently moved toward Claude Code is lower token consumption. Anything Cursor can do, Claude Code can handle too; for frontend developers in particular, Claude Code's generated output tends to align more closely with expectations.

This comparison reflects a broader trend in the AI coding tool landscape: cost of use and output quality are becoming the two most important factors developers weigh when choosing a tool.

Installing Claude Code: Three Steps to Get Ready

Claude Code supports multiple usage modes: terminal command line, IDE plugin integration, web browser, and app. For developers, the most recommended approach is the terminal.

Prerequisites: Git and Node

Before installing Claude Code, you'll need two tools:

1. Git: Download the Windows installer from the official site and install with default options. Once done, type git in your terminal — no errors means success.
2. Node: Download the version matching your OS from the official site. The newer 24.x release is recommended. After installation, run node -V — if it prints a version number, you're good.

Both installers are fairly small. It's recommended to keep the default C: drive installation path unless you're genuinely tight on space.

Running the Installation Command

With dependencies in place, open Windows PowerShell (no admin rights needed), copy the installation command from the Claude official website, paste it in, and wait a few minutes.

When the terminal outputs a version number (e.g., 2.1.104), installation is complete. The executable is typically located at C:\Users\YourUsername\.local\bin\. However — a successful install doesn't mean you can use it right away. You'll still need to complete the following configuration.

Configuring Claude Code: Solving Three Common Roadblocks

This is the most critical — and most frustrating — part of the process. Users in China typically need to work through three issues in sequence.

Issue 1: Setting the Environment Variable

Some users get a "not recognized as an internal or external command" error after typing claude in the terminal. This happens because the installation path wasn't added to the system environment variables.

Steps to fix:

1. Confirm the installation directory (the .local\bin path mentioned above);
2. Type "environment variables" in the Windows search bar and open "Edit the system environment variables";
3. Under User variables, find Path, double-click it, add a new entry with the full path to the bin directory, and save;
4. Restart PowerShell and type claude again — no error means it worked.

Issue 2: Bypassing the Connectivity Check

After fixing the environment variable, you may still encounter authorization or connection failures at runtime. This is caused by network conditions in China — even with a proxy enabled, you may hit the same issue.

The fix is to edit a configuration file: locate .claude.json in your C: drive user directory, open it with Notepad or any text editor, and add the bypass configuration code at the appropriate location (make sure the JSON remains valid — add a comma before any new entries), then restart and it should work normally.

Issue 3: Connecting to a Domestic LLM

Once Claude Code is running, you'll find that the built-in Claude models aren't available — unless you have an active Claude subscription. For non-subscribers, you can connect to a domestic large language model. Alibaba Cloud Model Studio (百炼) is a solid choice here, mainly because it offers a generous free tier.

The integration process has three steps, all done in CMD by setting environment variables:

1. Set the API Key: Create an API Key in the Alibaba Cloud Model Studio console and write it to your config;
2. Set the API endpoint: Enter the public API base URL for Alibaba Cloud Model Studio;
3. Specify the model name: Choose from the free model list, such as Qwen or GLM series.

After all three steps, verify the output matches what you entered. Once everything checks out, Claude Code is fully ready to use.

Hands-On Demo: Generating a Web Page and Three Working Modes

With configuration complete, let's look at a concrete example to get a feel for what Claude Code can actually do.

Set Up a Workspace and Generate a Web Page with One Sentence

Good practice: always create a dedicated directory for your project first. Use mkdir on your F: drive to create a folder, navigate into it, run claude, choose to trust the current folder, and then issue a natural language command — something like "create a product landing page."

During generation, Claude Code will proactively ask clarifying questions about missing details (page style, featured functionality, etc.) before proceeding. The final output is a webpage with tech-inspired animations and solid visual polish — something that would realistically take a frontend developer a full day to build by hand.

Three Working Modes for Every Scenario

Press Shift + Tab to switch freely between three modes:

1. Default mode: Asks for your confirmation before creating or modifying any file — the most cautious and safest option;
2. Auto mode: Once authorized, file operations in the current session proceed automatically without interruption — the most efficient option;
3. Plan mode: Discussion only, no file modifications — ideal for complex requirements. Use it to have multi-turn conversations with the AI until it fully understands the business logic, confirm the approach, then switch to execution mode to actually implement it.

The core value of Plan mode: on complex projects, diving straight into execution often leads to problems. Aligning on the plan first and executing second significantly reduces errors and keeps results more predictable.

Summary

Claude Code represents a meaningful evolution in AI coding tools — from "offering advice" to "autonomous execution." For developers in China, the installation itself isn't hard; the real challenge lies in the three configuration hurdles: environment variables, bypassing the connectivity check, and connecting to a domestic LLM. Once you clear those, combined with three flexible working modes, Claude Code can genuinely become a powerful tool for boosting your development productivity.

If you're torn between Cursor and Claude Code, the best approach is to try both yourself — evaluate on token cost and actual output quality — then decide.