VSCode + Claude Code + DeepSeek: Complete Guide to Setting Up an AI Programming Environment

Why Do Embedded Developers Need an AI Programming Assistant?

We're living in an "AI Everywhere" era. Whether it's office automation, data analysis, software development, or creative design, AI is gradually becoming everyone's super assistant. For the embedded development field, this trend is equally impossible to ignore.

In the past, writing code was entirely self-reliant—memorizing various register configurations and API calls by rote. Now with AI programming assistants, they can handle repetitive work, suggest best practices, and even help you discover hidden bugs. As a Bilibili content creator mentioned in their ZYNQ development tutorial series: "Those who master tools outperform those who rely on memory." Learning to use AI not only makes development more effortless but also lets you focus on creative and strategic work.

It's worth noting that embedded development is fundamentally different from web or application development—it involves hardware register operations, timing constraints, interrupt handling, and other low-level details. Developers need to simultaneously understand hardware datasheets and software architecture. AI programming assistants are particularly valuable in this domain because they can quickly retrieve and synthesize critical information scattered across hundreds of pages of technical documentation, compressing what would take hours of manual document review into just seconds.

This article provides a detailed walkthrough of how to set up an AI-assisted programming environment from scratch based on VSCode + Claude Code + DeepSeek, helping embedded developers get started quickly.

Step 1: Install Git

Git is a prerequisite dependency for Claude Code and a fundamental tool for modern software development. Git is more than just a version control tool—it has become the infrastructure layer of modern software development. In embedded development, Git's value manifests in multiple dimensions: First, hardware driver code often requires maintaining multiple branches for different chip versions, and Git's branch management capabilities make this controllable. Second, Claude Code relies on Git to understand a project's file structure and change history, enabling more precise context-aware suggestions. Finally, Git's diff functionality allows developers to clearly see the differences between AI-generated code and existing code, facilitating review and rollback.

The installation steps are as follows:

Download and Installation

1. Open your browser and search for "Git," then navigate to the official Git website (git-scm.com)
2. Find "Install for Windows" and click to enter the download page
3. Among the various installation methods, select Standalone Installer
4. Click the Windows x64 version to download
5. After downloading, run the exe file. During installation, it's recommended to use the default configuration—just click "Next" all the way through

Verify Installation

After installation, open a cmd terminal and enter the following command:

git version

If the console successfully prints Git's version number, the installation was successful.

Step 2: Install Claude Code

Claude Code is Anthropic's official command-line AI programming tool that enables code-level interaction with AI directly in the terminal. It supports code generation, code analysis, and project-level context understanding, making it the core component of an AI programming environment.

Unlike traditional AI chat interfaces, Claude Code adopts an Agent architecture that can proactively read project files, understand directory structures, execute terminal commands, and reason within the entire project context. This means when you ask a question about register configuration, it can not only provide a generic answer but also offer targeted suggestions based on the header file definitions and driver frameworks already in your project. Anthropic positions it as an "agentic coding tool"—a programming agent with autonomous action capabilities, rather than a simple Q&A chatbot. For example, it can automatically scan your Makefile to understand compilation configurations, or read device tree files to infer hardware topology.

Install via PowerShell

1. Open your browser and search for "Claude Code," then find the official Claude documentation
2. Navigate to the "Get Started" page
3. Among the various installation methods, find the command-line installation method for PowerShell and copy the command
4. Open PowerShell on your computer, paste and execute the command
5. Wait for the installation to complete (this process may be slow—patience is required)

Verify Installation

After installation, enter the following in the terminal:

claude version

If version information is successfully printed, the installation was successful.

Tip: If you encounter network issues during installation, you may need to configure a proxy or use a mirror source.

Step 3: Install CC Switch for Multi-Model Management

This is the most practically valuable part of the entire AI programming environment setup. Today, large language models are emerging one after another—Google's Gemini, OpenAI's GPT series, DeepSeek, and more. Developers often need to switch between different models. If you have to manually log out and reconfigure each time, efficiency drops dramatically.

Different large language models show significant performance differences on embedded development tasks. The DeepSeek series excels in code understanding and generation, with its V3 and R1 versions demonstrating strong logical reasoning capabilities when handling low-level C/C++ code. In contrast, Claude has advantages in long document comprehension and complex architecture design, while GPT-4o excels in multimodal scenarios (such as interpreting circuit diagram screenshots or waveform plots). This is precisely why multi-model management tools exist—embedded development involves multi-layered tasks from register-level operations to system architecture design, and no single model can cover all scenarios.

What is CC Switch?

CC Switch functions like an AI model control center, with its core value being:

• Quick switching: Switch between different AI model instances with one click, without manual login or configuration
• Unified management: Centrally manage API Keys, account information, and model preferences
• Lower barriers: Make multi-model collaboration simple and intuitive

Installation Steps

1. Search for "CC Switch" in your browser and find the corresponding GitHub repository
2. Find the "Download" section on the page and click "Releases"
3. Select the latest version and find the Windows installation package in Assets to download
4. Run the installer—keep the path and configuration at their defaults

Configure the DeepSeek Model

After installation, open CC Switch and follow these steps to configure DeepSeek:

1. Click the "+" button in the upper right corner to add a model
2. Select DeepSeek from the model list
3. Enter your pre-obtained API Key (keep the API format and authentication fields at their defaults)
4. Select the specific model you want to use in the model selection field
5. Click "Add"—the model will appear in the list
6. Click "Enable" to activate the model

Regarding API Key acquisition: DeepSeek's API uses a per-token billing model, with pricing that offers orders-of-magnitude cost advantages compared to OpenAI and Anthropic. Taking DeepSeek-V3 as an example, its input token price is approximately 1/20 to 1/50 of GPT-4o. For embedded developers, daily use scenarios include code completion, documentation Q&A, and bug analysis, with average monthly API consumption typically ranging from a few yuan to tens of yuan RMB. Developers need to register an account on the DeepSeek official platform (platform.deepseek.com), create an API Key, and after adding funds, can connect through CC Switch.

If you want to add other models later (such as GPT-4, Gemini, etc.), the process is exactly the same. This unified management approach allows developers to flexibly choose the most suitable AI model based on different task scenarios.

Step 4: Integrate Claude Code in VSCode

With all command-line tools ready, the final step is integrating Claude Code into your daily VSCode development environment.

Install the VSCode Extension

1. Open VSCode and go to the Extensions marketplace
2. Search for "Claude"
3. Find the "Claude Code for VS Code" extension and click Install
4. After installation, the Claude Code launch icon will appear in the left sidebar

Start Using the AI Programming Assistant

Click the Claude Code icon on the left sidebar. Since we've already configured the model in CC Switch, Claude Code will automatically log in based on the selected model without manual intervention.

Click "New Session" to create a new conversation session, and you can start interacting with AI. You can ask it to analyze code, generate functions, explain register configurations, or even make code modifications directly within the project context.

Thoughts on Toolchain Selection

This AI programming environment setup has several noteworthy design considerations:

Why choose DeepSeek over Claude's native model? DeepSeek has a clear advantage in cost-effectiveness. For the extensive code reading and comprehension tasks in embedded development, DeepSeek's performance is already excellent, while API call costs are significantly lower. Specifically, embedded projects frequently require AI to read thousands of lines of driver code or BSP (Board Support Package) source code to answer specific questions—in these high-input-token scenarios, DeepSeek's cost advantage is particularly pronounced.

Why use CC Switch for multi-model management? This reflects a "don't put all your eggs in one basket" strategy. Different models excel at different tasks, and CC Switch lets you switch at any time to find the AI assistant best suited for the current task. For example, you might use DeepSeek (which excels at code generation) when writing an AXI bus driver, then switch to Claude (which excels at long document analysis) when you need to understand a complex Xilinx Technical Reference Manual.

Why base this on VSCode rather than a dedicated IDE? VSCode's plugin ecosystem and lightweight nature make it an ideal platform for integrating AI programming tools. For development scenarios like ZYNQ that span hardware and software, VSCode's flexibility is particularly important. ZYNQ is a SoC platform from Xilinx (now AMD) that integrates ARM processors with FPGA programmable logic. This heterogeneous architecture means developers need to simultaneously master embedded software (C/C++), hardware description languages (Verilog/VHDL), and the interaction interfaces between them (AXI bus protocols, etc.). In traditional development, engineers need to frequently switch between multiple tools like Vivado and Vitis, while VSCode paired with an AI programming assistant can help understand cross-domain knowledge in a unified environment—for example, explaining AXI4-Lite register mapping relationships, or generating corresponding driver code frameworks based on hardware constraints.

Summary

The complete workflow for setting up an AI-assisted programming environment is: Git → Claude Code → CC Switch → VSCode Extension. The entire process isn't complex, but each step serves a necessary purpose. Git provides the version control foundation, Claude Code provides AI interaction capabilities, CC Switch solves the multi-model management pain point, and the VSCode extension integrates everything into the development workflow.

For embedded developers, the value of an AI programming assistant goes beyond writing code faster—it can help you understand complex hardware documentation, analyze low-level driver logic, and quickly pinpoint debugging issues. The tools are ready; the next step is using them to truly boost your development efficiency.

Key Takeaways

• Complete AI programming environment setup flow: Git → Claude Code → CC Switch → VSCode Extension—configuration done in four steps
• CC Switch serves as an AI control center, providing unified management of multiple LLM API Keys and switching, avoiding frequent manual logins
• Choosing DeepSeek balances cost-effectiveness, making it suitable for code reading and comprehension tasks in embedded development
• After Claude Code integration via VSCode extension, it automatically recognizes CC Switch configuration for seamless in-IDE AI interaction
• The core value of AI programming assistants for embedded developers lies in understanding hardware documentation, analyzing driver logic, and quickly pinpointing issues
• Different AI models have different strengths: DeepSeek excels at code generation and understanding, Claude at long document analysis, and GPT-4o at multimodal scenarios
• Heterogeneous SoC development like ZYNQ involves hardware-software co-design, and AI assistants can bridge cross-domain knowledge gaps in a unified environment