---

name: exa description: Search the web and extract content using Exa AI. Use this skill when searching for current information, researching topics, fetching content from URLs, finding code examples, or when the user needs real-time web data. Also use for competitive research, finding similar pages, or extracting structured content from websites.

Exa Web Search & Content Extraction

Real-time web search and content extraction powered by Exa AI.

Prerequisite: Set EXA_API_KEY environment variable.

API Quick Reference

Usage

Web Search

Copy# Basic search
curl -X POST 'https://api.exa.ai/search' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"query": "latest AI research", "text": true}'

# Deep search with more results
curl -X POST 'https://api.exa.ai/search' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"query": "React server components", "type": "deep", "numResults": 10, "text": true}'

# Fast search
curl -X POST 'https://api.exa.ai/search' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"query": "node.js version", "type": "fast"}'

Extract URL Content

Copycurl -X POST 'https://api.exa.ai/contents' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"urls": ["https://docs.example.com/api"], "text": true}'

Code Context

Copycurl -X POST 'https://api.exa.ai/context' \
  -H "x-api-key: $EXA_API_KEY" \
  -H 'Content-Type: application/json' \
  -d '{"query": "React useState hook examples", "tokensNum": 5000}'

Workflow

Think step-by-step when using Exa:

1. Choose the right endpoint based on the task:

   • Web research, searching topics → /search
   • Extracting content from specific URLs → /contents
   • Programming examples, code context → /context

2. Select search type for /search:

   • fast - Quick results, lower latency
   • auto - Balanced (default)
   • deep - Comprehensive, higher quality

3. Request content when needed using "text": true or "context": true

When to Use Each Endpoint

Search Types

Rules

• Current year is 2026 - Use this for date-relative queries (e.g., "latest", "recent", "this year")
• Use "type": "fast" for simple factual queries
• Use "type": "deep" for research requiring comprehensive results
• Add "text": true to get full page content in search results
• Add "context": true for LLM-optimized context strings
• Use /context endpoint for programming questions instead of /search
• Prefer /contents when you have a specific URL to extract

References

• Search - Web search parameters and options
• Crawl - URL content extraction
• Code - Programming context search