/quick-start

Quick start

Get free-claude-code running with NVIDIA NIM in under 5 minutes.

Quick start

Get free-claude-code running with NVIDIA NIM in under 5 minutes. This is the fastest path to using Claude Code for free.

Prerequisites

You need two things before starting:

  1. NVIDIA NIM API key: Get one at build.nvidia.com/settings/api-keys
  2. Claude Code installed: Follow the official Claude Code installation guide

Step 1: Install uv

uv is the Python package manager and runner that free-claude-code uses.

pip install uv

If uv is already installed, update it:

uv self update

Step 2: Clone and configure

git clone https://github.com/Alishahryar1/free-claude-code.git
cd free-claude-code
cp .env.example .env

Edit .env and add your NVIDIA NIM API key:

NVIDIA_NIM_API_KEY="nvapi-your-key-here"

MODEL_OPUS=
MODEL_SONNET=
MODEL_HAIKU=
MODEL="nvidia_nim/z-ai/glm4.7"

ENABLE_THINKING=true

The MODEL variable sets your fallback model. MODEL_OPUS, MODEL_SONNET, and MODEL_HAIKU map Claude’s model tiers to specific provider models. When empty, they fall back to MODEL.

Step 3: Start the proxy

In one terminal window:

uv run uvicorn server:app --host 0.0.0.0 --port 8082

You should see startup logs indicating the server is listening on port 8082.

Step 4: Run Claude Code through the proxy

In a second terminal window, set two environment variables and launch Claude Code:

macOS/Linux (Bash/Zsh):

ANTHROPIC_BASE_URL="http://localhost:8082" ANTHROPIC_AUTH_TOKEN="freecc" claude

Windows (PowerShell):

$env:ANTHROPIC_AUTH_TOKEN="freecc"; $env:ANTHROPIC_BASE_URL="http://localhost:8082"; claude

Important: Point ANTHROPIC_BASE_URL at the proxy root URL (http://localhost:8082), not http://localhost:8082/v1.

Claude Code will start normally. All API requests now route through your local proxy to NVIDIA NIM’s free tier.

Verify it works

Ask Claude Code to perform a simple task:

> Write a Python function that calculates the Fibonacci sequence

You should see:

  1. Claude responds with code
  2. The proxy logs show requests going to integrate.api.nvidia.com
  3. No Anthropic API credits are consumed

Next steps

  • Install as a package: Use uv tool install to avoid cloning the repo every time.
  • Try other providers: Configure OpenRouter for more model variety, or LM Studio for fully local operation.
  • Set up VSCode: Configure the Claude Code VSCode extension to use the proxy.
  • Enable voice notes: Install voice extras and use voice messages with Discord or Telegram bots.

Troubleshooting

Proxy won’t start: Check that port 8082 is not in use by another process. Change the port with --port 8083 if needed.

Claude Code still asks for login: You may need to complete the OAuth flow once. After that, the proxy handles all API calls.

Model not found errors: Verify your MODEL variable uses the correct format: nvidia_nim/organization/model-name. Check build.nvidia.com/explore/discover for available models.