Rechercher

PythonReactSEOPDF
GitHub Skill

grok-search

Use Grok AI (grok.com) to search and research topics via browser automation. Triggers on "use grok to search", "ask grok", "grok research", "search with grok", or when user wants real-time X/Twitter data analysis, trending topics, or AI-powered web research.

kaichen☆ 0 Étoiles du dépôt0 Forks du dépôt

Installation

Suivez les instructions du dépôt pour votre agent. Consultez les fichiers et dépendances avant l’installation.

Voir la source ↗Guides

Instructions

Grok Search

Search and research using Grok AI via browser automation with Claude in Chrome MCP.

Prerequisites

  • Claude in Chrome MCP extension installed and configured
  • Logged into grok.com in Chrome

Quick Start

Direct URL Query (Recommended)

Use URL parameters to submit queries directly without manual input:

https://grok.com/?q={your_query}

This automatically submits the query and starts generating results.

Example Workflow
// 1. Create new tab
mcp__claude-in-chrome__tabs_create_mcp()

// 2. Navigate with query parameter
mcp__claude-in-chrome__navigate({
  url: "https://grok.com/?q=your search query here",
  tabId: <tab_id>
})

// 3. Wait for results (typically 5-30 seconds for regular, 1-3 minutes for DeepSearch)
mcp__claude-in-chrome__computer({ action: "wait", duration: 10, tabId: <tab_id> })

// 4. Get page text content
mcp__claude-in-chrome__get_page_text({ tabId: <tab_id> })

// 5. Find and click copy button (aria-label="Copy") or (aria-label="复制") 
mcp__claude-in-chrome__find({ query: "复制 button", tabId: <tab_id> })
mcp__claude-in-chrome__computer({ action: "left_click", ref: "ref_xxx", tabId: <tab_id> })

// 6. Save to file via clipboard
Bash: pbpaste > result.md

Search Modes

Default Mode (Fast)
  • URL: https://grok.com/?q={query}
  • Response time: 5-15 seconds
  • Best for: Quick questions, simple lookups
DeepSearch Mode (Thorough)
  • Requires clicking DeepSearch button before submitting
  • Response time: 1-3 minutes
  • Best for: Research, comprehensive analysis, multi-source information

To enable DeepSearch:

// After navigating to grok.com (without query param)
mcp__claude-in-chrome__find({ query: "DeepSearch button", tabId: <tab_id> })
mcp__claude-in-chrome__computer({ action: "left_click", ref: "ref_xxx", tabId: <tab_id> })
// Then input query and submit

Common Use Cases

1. Real-time X/Twitter Analysis
Query: "24h内中文热门推特在讨论什么"
Query: "What are trending topics on X about AI in the last 24 hours"
Query: "24h内我关注的中文推特时间线里有哪些热门讨论"
2. News and Current Events
Query: "最近24小时内 Claude 相关的信息有哪些"
Query: "Latest news about [topic] in the past 24 hours"
3. Research Topics
Query: "[Topic] 的最新进展和讨论"
Query: "Comprehensive overview of [topic]"

Handling Results

Method 1: Copy Button (Preferred)
// Find copy button using aria-label
mcp__claude-in-chrome__find({ query: "复制 button", tabId: <tab_id> })
// Click to copy content to clipboard
mcp__claude-in-chrome__computer({ action: "left_click", ref: "ref_xxx", tabId: <tab_id> })
// Save from clipboard
Bash: pbpaste > result.md
Method 2: Direct Text Extraction (Fallback)
// If copy button fails, extract text directly
const text = mcp__claude-in-chrome__get_page_text({ tabId: <tab_id> })
// Then save content via Write tool

Tips

  1. Check Generation Status: Look for timing indicators like "思考了 Xs" or "Searching" in page text
  2. Wait Adequately: DeepSearch can take 1-3 minutes; check periodically with get_page_text
  3. Tab Management: Create new tabs for parallel searches
  4. Error Recovery: If extension conflicts occur, try get_page_text as fallback for content extraction
  5. URL Encoding: Complex queries with special characters should be URL-encoded

Troubleshooting

Extension Conflicts

If you encounter "Cannot access a chrome-extension:// URL" errors:

  1. Try navigating to a fresh URL
  2. Use get_page_text instead of screenshot
  3. Create a new tab and retry
Results Not Loading
  1. Increase wait time
  2. Check if Grok is still generating (look for loading indicators)
  3. Verify you're logged into grok.com
Copy Button Not Working
  1. Use get_page_text to extract content directly
  2. Format and save the text content manually

Exemples

Consultez les exemples du dépôt source.

Voir la source ↗

Historique source

Date d’indexation
2026-01-20
Mise à jour de l’index
2026-01-26
Branch
main

Les dates concernent ce répertoire. Consultez le dépôt pour l’historique du code.

GitHub ↗

chat Comments (0)

chat_bubble_outline

No comments yet. Be the first to share your thoughts!