Exa Search Skill

This skill provides comprehensive access to the Exa AI search API through 5 specialized scripts, each targeting a specific endpoint with full parameter support.

When to Use This Skill

Use this skill when:

•Performing neural web searches with AI-powered relevance ranking
•Searching specific categories (research papers, companies, news, GitHub, people, etc.)
•Extracting clean content from known URLs
•Finding pages similar to a reference URL (competitive analysis, related research)
•Conducting AI-powered research with automatic source citations
•Building datasets from web content
•Performing domain-filtered searches for authoritative sources
•Need date-filtered results for recent information

Prerequisite

All scripts require the EXA_API_KEY environment variable. Get your key at https://dashboard.exa.ai

bash

# Verify API key is set
echo $EXA_API_KEY

Available Scripts

1. exa_search.py - Neural Web Search

Command: python3 ~/.claude/skills/exa-search/scripts/exa_search.py

The most comprehensive search script with full access to Exa's /search endpoint.

Search Types:

•auto (default) - Intelligently combines search methods
•neural - Pure embeddings-based semantic search
•fast - Lower latency, streamlined search
•deep - Comprehensive search with query expansion

Categories:

•company - Company pages and profiles
•research paper - Academic papers (arXiv, etc.)
•news - News articles
•pdf - PDF documents
•github - GitHub repositories and code
•tweet - Twitter/X posts
•personal site - Personal websites and blogs
•people - LinkedIn profiles and people pages
•financial report - Financial documents

Examples:

bash

# Basic search
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "AI agent frameworks"

# Category-filtered search
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "transformer architecture" --category "research paper" -n 20

# Deep search with multiple query variations
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "web scraping Python" --deep --additional-queries "Python crawler" "BeautifulSoup Scrapy"

# Domain-filtered search
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "Python tutorials" --domains docs.python.org realpython.com github.com

# Date-filtered search for recent content
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "OpenAI announcements" --after 2024-06-01 --category news

# Search with content for RAG applications
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "React hooks best practices" --context --context-chars 10000

# Search with summaries and highlights
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "startup funding" --category company --summary "funding history and investors" --highlights

# Fast search for quick lookups
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "JavaScript fetch API" --fast -n 5

# Search for people/profiles
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "machine learning engineers" --category people

# Search with subpage crawling
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "React documentation" --subpages 3 --domains react.dev

# Output as JSON
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "GraphQL vs REST" --json

Key Parameters:

•-n, --num - Number of results (max 100)
•--neural/--fast/--deep - Search type
•--category, -c - Filter by category
•--domains - Only include these domains
•--exclude-domains - Exclude these domains
•--after/--before - Date filtering (YYYY-MM-DD)
•--context - Combine results into RAG context string
•--summary - Generate AI summaries
•--highlights - Include key excerpts
•--subpages - Crawl subpages
•--json - Output raw JSON

2. exa_contents.py - URL Content Extraction

Command: python3 ~/.claude/skills/exa-search/scripts/exa_contents.py

Extract clean, formatted content from specific URLs using Exa's /contents endpoint.

Examples:

bash

# Basic content extraction
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py "https://arxiv.org/abs/2307.06435"

# Multiple URLs
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py URL1 URL2 URL3

# With AI summary
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py "https://docs.example.com" --summary "Key API methods"

# With highlights and subpages
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py "https://docs.python.org/3/tutorial" --highlights --subpages 5

# Livecrawl for fresh content
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py "https://news.ycombinator.com" --livecrawl always

# Extract links and images
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py "https://example.com" --links 10 --images 5

# Limit text length
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py "https://long-article.com" --max-chars 5000

# JSON output
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py URLs --json

Key Parameters:

•--summary - Generate AI summary with query
•--highlights - Extract key excerpts
•--subpages - Crawl linked subpages
•--livecrawl - Fresh content mode: always, preferred, fallback, never
•--context - Combine contents into RAG context string
•--context-chars - Limit context string length
•--links/--images - Extract links/images
•--max-chars - Limit text length
•--json - Output raw JSON

3. exa_similar.py - Find Similar Pages

Command: python3 ~/.claude/skills/exa-search/scripts/exa_similar.py

Discover pages semantically similar to a reference URL using Exa's /findSimilar endpoint.

Use Cases:

•Find similar research papers
•Discover competitor products/companies
•Expand reading lists with related content
•Competitive analysis
•Content recommendations

Examples:

bash

# Find similar pages
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://arxiv.org/abs/2307.06435"

# Find similar with more results
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://github.com/langchain-ai/langchain" -n 20

# Similar companies (exclude source domain)
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://stripe.com" --category company --exclude-source

# Similar from specific domains only
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://medium.com/article" --domains medium.com substack.com

# With comparison summaries
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://product.com" --summary "How is this different?"

# Recent similar content only
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://blog.example.com" --after 2024-01-01

# JSON output
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py URL --json

Key Parameters:

•-n, --num - Number of similar results (max 100)
•--category - Filter by category
•--domains - Only these domains
•--exclude-source - Exclude the source URL's domain
•--after/--before - Date filtering
•--must-include - Results must contain these strings
•--must-exclude - Results must NOT contain these strings
•--safe - Enable content moderation filter
•--context - Combine contents into RAG context string
•--context-chars - Limit context string length
•--summary - Generate comparison summaries
•--highlights - Key excerpts
•--json - Output raw JSON

4. exa_research.py - AI-Powered Deep Research

Command: python3 ~/.claude/skills/exa-search/scripts/exa_research.py

Conduct deep research with AI-synthesized answers and automatic citations using Exa's /answer endpoint.

Use Cases:

•Quick research on any topic with citations
•Fact-checking with source verification
•Technical documentation queries
•Market research summaries
•Competitive analysis
•Current events synthesis

Examples:

bash

# Basic research question
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "What are the main differences between React and Vue?"

# Research with source details
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "SpaceX latest achievements" --sources

# Streaming answer in real-time
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "Explain quantum computing" --stream

# Domain-filtered research (authoritative sources only)
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "Python async best practices" --domains docs.python.org realpython.com

# Recent information only
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "AI regulation news" --after 2024-06-01

# Markdown output with citations
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "Compare cloud providers" --markdown

# Answer only (for piping)
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "What is RLHF?" --answer-only

# With key excerpts from sources
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "Transformer architecture explained" --highlights --sources

# More sources for comprehensive research
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "GraphQL vs REST comparison" -n 10

# JSON output
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "query" --json

Key Parameters:

•-n, --num - Number of sources (default: 5, max: 20)
•--domains - Only use these domains as sources
•--after/--before - Date filtering
•--stream - Stream the answer in real-time
•--sources - Show detailed source information
•--highlights - Include key excerpts
•--markdown - Output as markdown with citations
•--answer-only - Only output the answer text
•--json - Output raw JSON

5. exa_research_async.py - Async Research with Pro Models

Command: python3 ~/.claude/skills/exa-search/scripts/exa_research_async.py

Advanced async research using Exa's /research/v1 endpoint with multiple models and structured output.

Features:

•Async job-based research (submit and poll)
•Model selection: exa-research-fast, exa-research, or exa-research-pro
•Structured JSON output with custom schemas
•Long-running research for complex topics
•Status tracking and job management

Examples:

bash

# Create research job
python3 ~/.claude/skills/exa-search/scripts/exa_research_async.py "What species of ant are similar to honeypot ants?"

# Use pro model and wait for completion
python3 ~/.claude/skills/exa-search/scripts/exa_research_async.py "Compare top 5 AI agent frameworks" --pro --wait

# Use fast model for quick tasks
python3 ~/.claude/skills/exa-search/scripts/exa_research_async.py "Quick market overview" --fast

# Structured output with JSON schema
python3 ~/.claude/skills/exa-search/scripts/exa_research_async.py "Analyze AI startup funding" \
  --schema '{"startups": [{"name": "string", "funding": "number", "focus": "string"}]}'

# Check job status
python3 ~/.claude/skills/exa-search/scripts/exa_research_async.py status r_abc123

# List all research jobs
python3 ~/.claude/skills/exa-search/scripts/exa_research_async.py list --limit 20

Key Parameters:

•--fast - Use exa-research-fast model (quicker/cheaper)
•--pro - Use exa-research-pro model (enhanced synthesis)
•--schema - JSON schema for structured output
•--wait - Wait for job completion
•--timeout - Max seconds to wait (default: 300)
•list - List all research jobs
•status <id> - Check status of a job

Script Selection Guide

Task	Best Script	Example
Web search with filters	`exa_search.py`	Find recent AI papers
Research papers	`exa_search.py --category "research paper"`	Academic literature
Company/startup info	`exa_search.py --category company`	Market research
GitHub repos	`exa_search.py --category github`	Code discovery
Extract URL content	`exa_contents.py`	Known URLs
Find competitors	`exa_similar.py --exclude-source`	Competitive analysis
Related research	`exa_similar.py`	Expand literature
Quick answers	`exa_research.py`	Fact-finding
Research with citations	`exa_research.py --sources`	Report writing
Complex structured research	`exa_research_async.py --pro`	Multi-source synthesis
RAG context	`exa_search.py --context`	LLM applications

Comparison with Exa MCP Tools

Feature	MCP Tools	This Skill's Scripts
Web Search	`mcp__exa__web_search_exa`	`exa_search.py` (more options)
Code Context	`mcp__exa__get_code_context_exa`	`exa_search.py --category github`
Contents Extraction	N/A	`exa_contents.py`
Find Similar	N/A	`exa_similar.py`
Quick Answer	N/A	`exa_research.py`
Async Research (Pro)	N/A	`exa_research_async.py`
Deep Search	N/A	`exa_search.py --deep`
Category Filtering	Limited	All 9 categories
Date Filtering	Limited	Full support
Domain Filtering	Limited	Full support
Subpage Crawling	N/A	Full support
Livecrawling	N/A	Full support
Structured Output	N/A	Full support

This skill provides full Exa API access beyond what MCP tools offer.

Cost Considerations

Exa API pricing (approximate):

•Neural search (1-25 results): $0.005
•Neural search (26-100 results): $0.025
•Deep search (1-25 results): $0.015
•Deep search (26-100 results): $0.075
•Content text: $0.001/page
•Highlights: $0.001/page
•Summary: $0.001/page

To optimize costs:

•Use --fast for quick lookups
•Limit results with -n
•Use --no-text if content not needed
•Filter by domains to reduce irrelevant results

Common Workflows

Workflow 1: Research a Topic

bash

# Quick research with citations
python3 ~/.claude/skills/exa-search/scripts/exa_research.py "How does RAG work?" --sources --markdown

Workflow 2: Literature Review

bash

# Find research papers
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "transformer optimization techniques" \
  --category "research paper" -n 20 --summary "Key contributions"

# Find similar papers to a key reference
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://arxiv.org/abs/1706.03762" \
  --category "research paper" -n 15

Workflow 3: Competitive Analysis

bash

# Find similar companies
python3 ~/.claude/skills/exa-search/scripts/exa_similar.py "https://competitor.com" \
  --category company --exclude-source -n 10 --summary "Key differentiators"

Workflow 4: Documentation Research

bash

# Search docs with domain filtering
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "React useEffect cleanup" \
  --domains react.dev developer.mozilla.org --context

# Extract specific docs content
python3 ~/.claude/skills/exa-search/scripts/exa_contents.py "https://react.dev/reference/react/useEffect" \
  --highlights --subpages 2

Workflow 5: News Monitoring

bash

# Recent news search
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "AI regulation" \
  --category news --after 2024-06-01 -n 15

Workflow 6: Build RAG Context

bash

# Get context string for LLM
python3 ~/.claude/skills/exa-search/scripts/exa_search.py "Python async programming patterns" \
  --context --context-chars 15000 --domains docs.python.org realpython.com

Test Suite

Run the test suite to verify API connectivity and feature functionality:

bash

# Run all tests
python3 ~/.claude/skills/exa-search/scripts/test_exa.py

# Run quick validation only
python3 ~/.claude/skills/exa-search/scripts/test_exa.py --quick

# Show detailed output
python3 ~/.claude/skills/exa-search/scripts/test_exa.py --verbose

# Test specific endpoint
python3 ~/.claude/skills/exa-search/scripts/test_exa.py --endpoint search
python3 ~/.claude/skills/exa-search/scripts/test_exa.py --endpoint contents
python3 ~/.claude/skills/exa-search/scripts/test_exa.py --endpoint similar
python3 ~/.claude/skills/exa-search/scripts/test_exa.py --endpoint answer
python3 ~/.claude/skills/exa-search/scripts/test_exa.py --endpoint research

The test suite verifies:

•API key validation
•Basic search, category filtering, domain filtering, deep search
•Content extraction, livecrawl, context parameter
•Find similar pages, moderation filter
•Research/answer endpoint, streaming support
•Async research model validation, job listing