How to Install Claude Code Free with Nvidia NIM Step by Step (No GPU, No Paid)

What if you could harness the exact same agentic workflow entirely for free?

Believe it or not, getting high-level AI capabilities for zero cost is 100% real, no complex cheats, and absolutely no expensive local GPU required. By leveraging an open-source local proxy tool called free-claude-code (developed by Ali Shahryar), you can intercept your local Claude Code traffic and safely route it to NVIDIA NIM, which provides free, high-speed access to top-tier open-source models with a generous rate limit of 40 requests per minutue.

In this comprehensive guide, we will walk you through setting up this architecture step-by-step using Windows PowerShell, configuring your local proxy environment, and unlocking hidden configuration features to swap models effortlessly. Let's dive in!

Step 1: Install Claude Code for Windows

To route Claude Code through our local proxy, we first need to install the official command-line interface (CLI) on our machine.

1. Navigate to the official Claude docs website. Find the command or
2. Open a standard Windows PowerShell window (you don't even need to run it as Administrator) and paste the following network command to initiate the secure download:

irm https://claude.ai/install.ps1 | iex

Step 2: Install UVs

To host our custom backend proxy and process the data payloads going back and forth, we need a fast, modern environment setup. For this project, we will use uv the blazing-fast Python package and environment installer written in Rust alongside Python 3.14.

powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
uv self update
uv python install 3.14

Step 3: Register Nvidia NIM API

Now that your local machine has the runtime requirements installed, we need to secure the "brains" of our operation. Instead of using Anthropic's paid infrastructure, we will use NVIDIA NIM (NVIDIA Inference Microservices), which lets us access incredibly fast, open-source LLMs hosted on NVIDIA's enterprise GPUs completely for free.

Follow these steps to grab your free API key:

1. Register for a Account Build Nvidia
2. Click on the Login or Join button in the top right corner.
3. Sign up using your email or an existing Google/GitHub account. The registration process is straightforward, completely free, and doesn't require any credit card details.
4. Once you log into your dashboard, click on your Profile icon in the top right corner to review your current developer tier limits.
5. Click on your profile dropdown menu, click on API Keys.
6. Click the green Generate API Key button.

Important: Copy the generated key (nvapi-...) immediately and save it in a temporary text file. For security reasons, NVIDIA will only show you this key once, and it will be hidden as soon as you close the popup window.

Step 4: Install and Launch the Proxy Server

With your tools ready and your NVIDIA API key generated, it’s time to pull down the repository that glues everything together: free-claude-code. We will use uv to install this project directly as a global command-line tool.

1. Open your PowerShell terminal and execute the following uv command to fetch, build, and install the proxy server straight from the GitHub source:

   uv tool install --force git+https://github.com/Alishahryar1/free-claude-code.git

2. Once the installation finishes, you can start the proxy server right from any terminal window by running: 

   fcc-server

   As soon as the engine successfully boots up, look closely at your PowerShell console output. The local network links will be printed cleanly on your screen: 

   Server URL: http://127.0.0.1:8082
   Admin UI: http://127.0.0.1:8082/admin

 Step 5: Configure NVIDIA NIM in the Admin UI

Now that your local server is running, you need to link it to your NVIDIA account so it can handle your coding requests. 

1. Open the Admin UI URL printed in your terminal usually localhost admin
2. Locate the text box designated for the API key.
3. Paste your NVIDIA API Key (the nvapi-... string you copied during Step 3) directly into the field.
4. Then copy the model to the routing model and click Apply to save the setting file.

Step 6: Configure Claude Settings and Run

To make the official Claude Code CLI communicate with our local proxy instead of hitting Anthropic’s paid servers, we need to declare the environment variables. While you can type these variables manually every time you launch the terminal, it is much easier to save them directly inside your Claude Global Settings file so they load automatically.

Claude Code stores its system configuration inside your Windows user directory.

1. Press Win + R, type %USERPROFILE%\.claude and press Enter. (Alternatively, navigate to C:\Users\YourUsername\.claude\).

2. Inside this folder, look for a file named settings.json. If it doesn't exist, create a new text file and rename it exactly to settings.json.

3. Open the file using Notepad or VS Code.

4. Add the Proxy Redirection Rules: Paste the following JSON block into the file. If you already have settings configured, append the "env" object inside your root brackets. 

   {
     "env": {
       "ANTHROPIC_BASE_URL": "http://localhost:8082",
       "ANTHROPIC_AUTH_TOKEN": "freecc",
       "CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY": "1"
     }
   }

With your settings saved and your proxy terminal console running quietly in the background, open a new Windows PowerShell session, navigate to any coding project folder, and fire up the agent. The Claude Code terminal interface will initialize instantly. Your commands will now run under a high-speed, free NVIDIA NIM-hosted architecture with zero billing worries!