grok-cli

name	grok-cli
description	CLI interface to Grok AI. Send prompts and get responses via command line, enabling integration with Claude Code and other tools. Uses stealth browser with Chrome auth for authenticated access.

Grok CLI Skill

Query Grok AI from the command line using your existing X.com authentication. Perfect for integrating Grok responses into Claude Code workflows.

When to Use This Skill

Trigger when user:

• Wants to query Grok/xAI
• Mentions "ask Grok", "prompt Grok", "Grok says"
• Needs real-time information that Grok has access to
• Wants to compare responses between Claude and Grok
• Needs X.com/Twitter-specific information

Prerequisites

1. Logged into X.com in Chrome: The skill uses your Chrome session
2. Grok access: Must have access to Grok on X.com (Premium subscription)

Critical: Always Use run.py Wrapper

NEVER call scripts directly. ALWAYS use python scripts/run.py [script]:

# CORRECT:
python scripts/run.py grok.py --prompt "Your question here"

# WRONG:
python scripts/grok.py --prompt "..."  # Fails without venv!

Core Usage

Basic Query (uses grok.com, requires --show-browser)

python scripts/run.py grok.py --prompt "What is the latest news about AI?" --show-browser

Use x.com/i/grok (alternative endpoint)

python scripts/run.py grok.py --prompt "Hello" --xcom

With Options

# Longer timeout for complex queries
python scripts/run.py grok.py --prompt "Explain quantum computing" --timeout 120 --show-browser

# Save screenshot
python scripts/run.py grok.py --prompt "What's trending?" --screenshot /tmp/grok.png --show-browser

# JSON output for parsing
python scripts/run.py grok.py --prompt "Capital of France?" --json --show-browser

# Raw output (just the response text, for piping)
python scripts/run.py grok.py --prompt "One word answer: 2+2=" --raw --show-browser

# Use different model (grok-2 has higher rate limits)
python scripts/run.py grok.py --prompt "Hello" --model grok-2 --show-browser

# Run multiple queries in parallel
python scripts/run.py grok.py --prompt "Query 1" --show-browser --session-id a &
python scripts/run.py grok.py --prompt "Query 2" --show-browser --session-id b &
wait

Piping to Other Tools

# Get Grok's response and pipe it
python scripts/run.py grok.py --prompt "List 5 trending topics" --raw | head -5

# Use in shell scripts
GROK_RESPONSE=$(python scripts/run.py grok.py --prompt "What day is it?" --raw)
echo "Grok says: $GROK_RESPONSE"

Script Reference

grok.py - Main Prompt Interface

python scripts/run.py grok.py --prompt "..." [options]

Options:
  --prompt, -p    The prompt to send to Grok (required)
  --timeout, -t   Response timeout in seconds (default: 60)
  --screenshot    Save screenshot to this path
  --show-browser  Show browser window for debugging
  --json          Output full JSON response with metadata
  --raw           Output only response text (no formatting)

Output Formats

Default (formatted)

============================================================
Prompt: What is 2+2?
============================================================

4

============================================================

JSON (--json)

{
  "success": true,
  "response": "4",
  "prompt": "What is 2+2?",
  "cookies_used": 20
}

Raw (--raw)

4

Integration with Claude Code

Use this skill when Claude Code needs information that:

• Requires real-time data (news, trends, current events)
• Is specific to X.com/Twitter ecosystem
• Would benefit from a second AI perspective

Example workflow in Claude Code:

# Claude Code can invoke this to get Grok's take
python ~/.claude/skills/grok-cli/scripts/run.py grok.py \
  --prompt "What are the latest developments in AI safety?" \
  --raw

Environment Management

The virtual environment is automatically managed:

• First run creates .venv automatically
• Dependencies install automatically
• Everything isolated in skill directory

Manual setup (only if automatic fails):

cd ~/.claude/skills/grok-cli
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Data Storage

All data stored in ~/.claude/skills/grok-cli/data/:

• screenshots/ - Saved screenshots
• browser_profile/ - Browser state for stealth

Configuration

Optional .env file:

HEADLESS=true              # Default browser visibility
DEFAULT_TIMEOUT=60         # Response timeout (seconds)

Troubleshooting

Problem	Solution
"Cloudflare challenge detected"	Use --show-browser flag (required for grok.com)
"Cookie extraction failed"	Login to X.com in Chrome
"Authentication failed"	Re-login to X.com, cookies may have expired
"Grok is under heavy usage"	Try again later, or sign in to grok.com for priority
"Rate limit reached"	Wait for reset or use --model grok-2 for higher limits
"Could not find input field"	UI may have changed, try --xcom flag
ModuleNotFoundError	Use run.py wrapper
Timeout waiting for response	Increase --timeout, try --show-browser

Limitations

• macOS only - Cookie decryption uses macOS Keychain
• --show-browser required for grok.com - Cloudflare blocks headless mode
• Rate limits - Thinking: 15/20hrs (Premium), Grok-2: higher limits, Premium+: unlimited
• Capacity limits on grok.com - Guest access may be throttled during high traffic
• No conversation history (each prompt is fresh)

How It Works

1. Cookie Extraction: Reads X.com cookies from Chrome's database
2. Decryption: Decrypts encrypted cookie values using macOS Keychain
3. Stealth Browser: Launches nodriver (undetected Chrome)
4. Cookie Injection: Sets cookies via Chrome DevTools Protocol
5. Navigation: Opens Grok on X.com
6. Interaction: Types prompt, sends, waits for response
7. Extraction: Polls for stable response text, returns it