Claude Code From Zero to Production: Installation, Switching to Chinese LLMs, and Git Workflow Guide

Claude Code (referred to as CC throughout this article) is one of the most talked-about AI programming tools available today. It has evolved from a simple command-line code analysis utility into a full-fledged "real development partner" with genuine engineering capabilities. This guide draws from a comprehensive hands-on tutorial to systematically walk you through CC's core capabilities, environment setup, switching to Chinese domestic models, and Git workflow integration — helping beginners get up to speed without the usual pitfalls.

From Code Analysis Tool to Engineering Collaborator

Claude Code started as a CLI-only tool, limited to analyzing project code, understanding file relationships, and suggesting modifications. After rapid iteration, today's CC looks nothing like its early form.

The core capability upgrades span three dimensions:

• Context awareness: Evolved from understanding single files to querying an entire project structure and delivering deep analysis.
• Engineering focus: No longer just a development aid — CC can now implement complete engineering workflows like SDD (Specification-Driven Development) and automated testing.
• Customizability: The Skill mechanism lets teams codify real development experience into reusable capabilities, significantly boosting team efficiency.

In short, CC has shifted from being a "tool for programming tasks" to a "true collaborator." Humans set goals, define constraints, and make judgment calls; AI handles execution, analysis, and comparison — absorbing the repetitive work.

Value for Different Roles

For programming beginners, CC can help read complex open-source code, pinpoint project errors and production issues, and even surface historical bugs and performance risks while helping fix them.

For independent developers, CC can orchestrate multiple agents simultaneously — acting as product manager, developer, and tester all at once — dramatically cutting the time cost of full-stack development.

For engineering teams, CC's greatest value lies in standardizing practices and reducing collaboration friction. Using CLAUDE.md config files to enforce team coding conventions, and building custom Skills for code review, issue analysis, and standards checking, new team members can onboard quickly even without complete project documentation.

Environment Setup: CLI Installation and Version Selection

CC is currently available in three main forms:

1. Web version: Access directly in a browser, best for lightweight conversational use cases;
2. CLI (command line): The original form — core commands are still worth knowing;
3. IDE integration: Integrated via plugins into VS Code, Cursor, or PyCharm/IDEA — the recommended approach for efficient day-to-day use.

IDE integration is recommended as your primary workflow, but understanding the underlying CLI commands remains important — the visual interfaces are ultimately just wrappers around those same commands.

CLI Installation Steps

CLI installation is straightforward and completed with a single command. Mac / Linux / WSL users should use the official one-click install command; Windows users use the corresponding instruction. After installation, run claude -v to check the version number — a successful output confirms the installation worked.

Node.js must be configured before installing CC. Download and install it from the Node.js official website, then run node -v to confirm the version. If you run into network issues, you can switch npm to a domestic Chinese mirror to speed up dependency downloads.

To update CC later, simply run claude update. CC will also automatically prompt you when a new version is available. For first-time installation, the official one-click method is recommended over installing directly via npm — it offers better stability.

Switching to Chinese LLMs: CC Switch Usage Guide

Due to network access issues, CC's default official models are often unreliable in China. The recommended solution is CC Switch, a tool for switching to domestic Chinese LLMs — currently the most widely adopted approach.

Here's how to set it up:

1. Download the latest version from the CC Switch project page. You can choose either the installer (MSI) or the portable version. Windows users should note whether they need the ARM or AMD64 build.
2. Open CC Switch and obtain an API key for your target model. DeepSeek and Kimi are popular choices; Tongyi Qianwen, Zhipu GLM, and others are also supported.
3. Click the "+" button to add a model, enter your API key (other address fields can remain at their defaults), then click "Enable."
4. In settings, check "Local Routing," then restart CC for the changes to take effect.

Once configured, use the slash command /model in the chat window to view and select from your configured domestic models. The experience with Chinese domestic models is largely the same as with the official model — differences show up mainly in output quality. Given the stability advantage, using a domestic model often delivers a smoother day-to-day experience.

CLI Commands and Getting Started Tips

Familiarity with the basic commands is a prerequisite for using Claude Code effectively. You can use help to see the full list of commands, or simply ask CC in natural language — for example, "Which command should I use to restore my last conversation?" — and it will give you the answer directly, no need to dig through official documentation.

Quick reference for commonly used commands:

• claude -c / --continue: Resume the last conversation (used most frequently);
• claude -r / --resume: View conversation history and selectively restore a previous session;
• claude -p: Headless mode — invoke CC's capabilities via a single command without entering an interactive session (e.g., output a project structure directly from the command line);
• /clear: Clear the current conversation and start a fresh session.

Typical First Steps

For beginners, one of the most common first use cases is having CC analyze an unfamiliar project structure. Navigate to the project directory and type something like "Help me analyze this project" — CC will automatically identify the tech stack and break down each module. This is especially useful for developers who've just taken over a legacy codebase.

You can also ask CC to implement specific functionality (like bubble sort). After writing the code, CC runs tests automatically. The standard modification workflow is: locate relevant files → show proposed changes → wait for user approval → apply changes — with support for approving items individually or accepting all at once.

Git Workflow in Practice: Fully Conversational Operations

Deep Git integration is one of Claude Code's most beloved features among developers. Users don't need to memorize complex Git commands — just describe what you need in natural language, and CC will automatically invoke the corresponding commands to get it done.

Here's a complete demonstration of a fully conversational Git workflow:

1. Create a project: "Help me create a personal tech blog using only HTML + CSS + JS";
2. Pull and push: "Help me pull the code" → "Move the files to the Test directory, commit and push" — CC automatically runs the corresponding Git commands;
3. Query commit history: "How many commits does the Test project have? Show each commit description in a table";
4. Branch-based development: "Create a dev branch, implement three different sorting algorithms, commit and push" — CC automatically creates the branch, writes the code, runs tests (all 12 pass), and pushes.

Automated Bug-Fix Loop

A more advanced use case is having CC complete the full loop of "identify problems → document problems → fix problems." Here's how: first ask CC to analyze the current branch for issues and output the results to issues.md (covering correctness problems, documentation-implementation inconsistencies, design consistency flaws, etc.). Then give the instruction: "Fix all issues based on issues.md, commit and push the code."

CC's process looks like: locate relevant code → understand context and root cause → propose and display a solution → run tests to verify the fix. No manual code edits required — the entire workflow is driven purely through conversation.

Summary

Claude Code has evolved from a code analysis tool into a genuine engineering collaborator. The core path to mastering this practical workflow is: set up Node.js and complete the CLI installation → use CC Switch to switch to a stable domestic model → get comfortable with basic CLI commands and interaction patterns → migrate your Git workflow to fully conversational operations.

"Let humans focus on decisions, let AI handle execution" — this isn't just a slogan. It's a working style that Claude Code can genuinely deliver in everyday development. For individuals and teams looking to boost development efficiency, this onboarding path is a reliable starting point for AI-driven engineering.