xAI Grok Search

Search the web and X (Twitter) using xAI's Grok API with real-time internet access, citations, and optional image/video understanding.

When to Use This Skill

Use Web Search For:

• Current information from websites, news articles, documentation
• Real-time data (stock prices, weather, recent events)
• Research topics with up-to-date web sources
• Finding information from specific websites/domains
• Verifying current facts

Use X Search For:

• What people are saying on X/Twitter about a topic
• Trending discussions and social sentiment
• Real-time reactions to events
• Posts from specific X handles/users
• Recent social media activity within date ranges

Do NOT use for:

• Historical facts that won't change
• General knowledge already available
• Mathematical calculations
• Code generation
• Creative writing

Setup

Required Environment Variables

export XAI_API_KEY="your-xai-api-key-here"

Get your API key from: https://console.x.ai/

Usage

The agent will automatically choose the right tool based on the user's query:

User: "What's the latest news about AI regulation?"

→ Uses web_search

User: "What are people saying about OpenAI on X?"

→ Uses x_search

API Reference

Function: search_web

Search the web using xAI's Grok API.

Parameters:

• query (required): Search query string
• model (optional): Model to use (default: "grok-4-1-fast-reasoning")
• allowed_domains (optional): Array of domains to restrict search (max 5)
• excluded_domains (optional): Array of domains to exclude (max 5)
• enable_image_understanding (optional): Enable image analysis (default: false)

Returns:

• content: The search response text
• citations: Array of sources with url, title, and snippet
• usage: Token usage statistics

Function: search_x

Search X (Twitter) using xAI's Grok API.

Parameters:

• query (required): Search query string
• model (optional): Model to use (default: "grok-4-1-fast-reasoning")
• allowed_x_handles (optional): Array of X handles to search (max 10, without @)
• excluded_x_handles (optional): Array of X handles to exclude (max 10, without @)
• from_date (optional): Start date in ISO8601 format (YYYY-MM-DD)
• to_date (optional): End date in ISO8601 format (YYYY-MM-DD)
• enable_image_understanding (optional): Enable image analysis (default: false)
• enable_video_understanding (optional): Enable video analysis (default: false)

Returns:

• content: The search response text
• citations: Array of X posts with url, title, and snippet
• usage: Token usage statistics

Implementation

This skill uses the xAI Responses API (/v1/responses endpoint).

Web Search

async function search_web(options) {
  const { query, model = 'grok-4-1-fast-reasoning', 
          allowed_domains, excluded_domains, enable_image_understanding } = options;

  const tool = { type: 'web_search' };
  if (allowed_domains) tool.allowed_domains = allowed_domains;
  if (excluded_domains) tool.excluded_domains = excluded_domains;
  if (enable_image_understanding) tool.enable_image_understanding = true;

  const response = await fetch('https://api.x.ai/v1/responses', {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'Authorization': `Bearer ${process.env.XAI_API_KEY}`
    },
    body: JSON.stringify({
      model,
      input: [{ role: 'user', content: query }],
      tools: [tool]
    })
  });

  const data = await response.json();
  return { 
    content: data.output[data.output.length - 1].content,
    citations: data.citations 
  };
}

X Search

async function search_x(options) {
  const { query, model = 'grok-4-1-fast-reasoning',
          allowed_x_handles, excluded_x_handles, from_date, to_date,
          enable_image_understanding, enable_video_understanding } = options;

  const tool = { type: 'x_search' };
  if (allowed_x_handles) tool.allowed_x_handles = allowed_x_handles;
  if (excluded_x_handles) tool.excluded_x_handles = excluded_x_handles;
  if (from_date) tool.from_date = from_date;
  if (to_date) tool.to_date = to_date;
  if (enable_image_understanding) tool.enable_image_understanding = true;
  if (enable_video_understanding) tool.enable_video_understanding = true;

  const response = await fetch('https://api.x.ai/v1/responses', {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'Authorization': `Bearer ${process.env.XAI_API_KEY}`
    },
    body: JSON.stringify({
      model,
      input: [{ role: 'user', content: query }],
      tools: [tool]
    })
  });

  const data = await response.json();
  return { 
    content: data.output[data.output.length - 1].content,
    citations: data.citations 
  };
}

Examples

Web Search - Current Events

const result = await search_web({ 
  query: "latest AI regulation developments" 
});

Web Search - Specific Domains

const result = await search_web({
  query: "UN climate summit latest",
  allowed_domains: ["un.org", "gov.uk", "grokipedia.com"]
});

X Search - Social Sentiment

const result = await search_x({
  query: "new iPhone reactions opinions"
});

X Search - Specific Handles

const result = await search_x({
  query: "AI thoughts",
  allowed_x_handles: ["elonmusk", "cstanley"],
  from_date: "2025-01-01"
});

X Search - With Media

const result = await search_x({
  query: "Mars landing images",
  enable_image_understanding: true,
  enable_video_understanding: true
});

Best Practices

Web Search

• Use allowed_domains to limit to specific domains (max 5)
• Use excluded_domains for known bad sources (max 5)
• Cannot use both at same time
• Enable image understanding only when needed

X Search

• Use allowed_x_handles to focus on specific accounts (max 10)
• Use excluded_x_handles to filter noise (max 10)
• Cannot use both at same time
• Don't include @ symbol in handles
• Use ISO8601 date format: YYYY-MM-DD
• Media understanding adds API costs

Troubleshooting

"XAI_API_KEY not found"

export XAI_API_KEY="your-key-here"

Rate Limiting

• Implement exponential backoff
• Cache frequent queries

Poor Results

• Add domain/handle filters
• Make queries more specific
• Narrow date ranges

Slow Responses

Search queries using reasoning models (e.g. grok-4-1-fast-reasoning) can take 30-60+ seconds to return, especially when the model performs multiple web or X searches. If the search is lagging, inform the user that results are still loading and ask them to type "poll" to check for the completed response.

API Documentation

• Web Search: https://docs.x.ai/developers/tools/web-search
• X Search: https://docs.x.ai/developers/tools/x-search