Grok CLI
Use Grok CLI with NanoGPT to access Grok models and 50+ other models through a single API key.
NanoGPT provides an OpenAI-compatible endpoint that works with the Grok CLI. This lets you:
- Use Grok CLI with your NanoGPT credits
- Access Grok models without a separate xAI account
- Use any other NanoGPT model through the same CLI
Step 1: Install Grok CLI
Prerequisites: Node.js 18 or newer
# Install Grok CLI
npm install -g @vibe-kit/grok-cli
# Verify installation
grok --version
Note: If you encounter permission issues during installation, try using sudo (macOS/Linux) or running the command prompt as an administrator (Windows) to re-execute the installation command.
Get API Key
- Go to NanoGPT API
- Create an account or log in
- Copy your API key
Set Required Environment Variables
Grok CLI reads these environment variables:| Variable | Value | Description |
|---|
GROK_BASE_URL | https://nano-gpt.com/api/v1 | NanoGPT OpenAI-compatible endpoint |
GROK_API_KEY | sk-nano-YOUR-API-KEY | Your NanoGPT API key |
macOS & Linux
Windows PowerShell
Windows Cmd
export GROK_BASE_URL="https://nano-gpt.com/api/v1"
export GROK_API_KEY="sk-nano-YOUR-API-KEY"
$env:GROK_BASE_URL = "https://nano-gpt.com/api/v1"
$env:GROK_API_KEY = "sk-nano-YOUR-API-KEY"
set GROK_BASE_URL=https://nano-gpt.com/api/v1
set GROK_API_KEY=sk-nano-YOUR-API-KEY
sk-nano-YOUR-API-KEY with your actual API key. Open a new terminal for changes to take effect if you set persistent variables. Optional: Make It Persistent
Add the variables to your shell profile (for example, ~/.zshrc or ~/.bashrc):echo 'export GROK_BASE_URL="https://nano-gpt.com/api/v1"' >> ~/.zshrc
echo 'export GROK_API_KEY="sk-nano-YOUR-API-KEY"' >> ~/.zshrc
source ~/.zshrc
Step 3: Start Using Grok CLI
Launch Grok CLI with a specific model:
Since you are connected to NanoGPT, you can use any available model:
# Grok models
grok --model grok-3
grok --model grok-3-fast
grok --model grok-3-mini
# Other providers through NanoGPT
grok --model openai/gpt-5.2
grok --model anthropic/claude-opus-4.5
grok --model google/gemini-3-flash-preview
Available Grok Models
| Model | Description |
|---|
grok-3 | xAI flagship model |
grok-3-fast | Faster, more cost-effective version |
grok-3-mini | Lightweight version |
Known Limitations
Grok CLI has limited compatibility with thinking or reasoning models. When you use a thinking model, the full reasoning content may be shown in the output.
If you want first-party CLI support for NanoGPT, consider Claude Code or Codex CLI.
Troubleshooting
| Issue | Solution |
|---|
401 Unauthorized | Check your API key is correct and has credits |
GROK_BASE_URL not set | Ensure the environment variable is exported |
Model not found | Verify the model name at nano-gpt.com/pricing |
Command not found: grok | Run npm install -g @vibe-kit/grok-cli again |
| Thinking content flooding output | Use a non-thinking model variant |
FAQ
Is Grok CLI official xAI software?
No. Grok CLI is a third-party package. NanoGPT works with it by providing an OpenAI-compatible endpoint.
Can I access Grok models from other clients?
Yes. You can use Grok models from Claude Code, Cursor, or any OpenAI-compatible client by setting the base URL to NanoGPT.
API Details (Reference)
| Setting | Value |
|---|
| Base URL | https://nano-gpt.com/api/v1 |
| Chat Completions Endpoint | POST /chat/completions |
| Models Endpoint | GET /models |
| Auth Header | Authorization: Bearer sk-nano-... |
Example cURL (verify connection)
curl -X POST "https://nano-gpt.com/api/v1/chat/completions" \
-H "Authorization: Bearer sk-nano-YOUR-API-KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "grok-3-fast",
"messages": [{"role": "user", "content": "Hello!"}]
}'
