majiayu000 Exa Research
AI-powered web search and content retrieval using Exa API.
Scripts Location
./scripts/
├── exa-search # Search the web
├── exa-contents # Get page contents
└── exa-similar # Find similar links
Core Workflows
1. Basic Search
./scripts/exa-search "your query"
Search Types:
auto(default) - Intelligently combines methodsfast- Quick results, lowest latencydeep- Comprehensive with query expansionneural- AI semantic search
Examples:
# Quick search
exa-search "React 19 new features"
# Deep research
exa-search "state of LLM agents 2025" --type deep --text
# News only
exa-search "OpenAI GPT-5" --category news --num 5
# Domain filtered
exa-search "machine learning tutorials" --include-domains arxiv.org,paperswithcode.com
2. Get Page Contents
./scripts/exa-contents "https://example.com" [options]
Examples:
# Get text content
exa-contents "https://blog.example.com/post" --text
# Get summary
exa-contents "https://paper.example.com" --summary
# Multiple URLs with LLM context
exa-contents "https://url1.com" "https://url2.com" --context
# Crawl subpages (for docs)
exa-contents "https://docs.example.com" --subpages 5
3. Find Similar Links
./scripts/exa-similar "https://example.com" [options]
Examples:
# Find competitors
exa-similar "https://stripe.com" --num 10
# Similar articles with content
exa-similar "https://blog.example.com/ai-post" --text
# Filter by domain
exa-similar "https://openai.com" --include-domains techcrunch.com,wired.com
Common Options Reference
| Option | Scripts | Description |
|---|---|---|
-t, --type |
search | auto, fast, deep, neural |
-n, --num |
search, similar | Number of results (max 100) |
-c, --category |
search | company, news, pdf, github, etc. |
--include-domains |
search, similar | Whitelist domains |
--exclude-domains |
search, similar | Blacklist domains |
--start-date |
search | Filter by publish date |
--text |
all | Include full page text |
--context |
all | Return LLM-optimized context |
--summary |
contents | AI-generated summary |
--highlights |
contents | Extract key passages |
--livecrawl |
contents | never, fallback, preferred, always |
Research Workflow Patterns
Pattern 1: Topic Research
# 1. Broad search with deep mode
exa-search "quantum computing applications 2025" --type deep --text | jq '.results[:5]'
# 2. Get detailed content for top results
exa-contents "https://result-url.com" --summary --context
Pattern 2: Competitive Analysis
# 1. Find similar companies
exa-similar "https://target-company.com" --num 10
# 2. Deep dive on specific competitor
exa-search "CompetitorName product features" --category company --text
Pattern 3: News Monitoring
# Recent news only
exa-search "AI regulation" --category news --start-date "2025-01-01" --num 20
Pattern 4: Technical Documentation
# Search for docs
exa-search "Next.js App Router caching" --include-domains nextjs.org,vercel.com --text
# Or crawl docs site
exa-contents "https://nextjs.org/docs/app/caching" --subpages 3 --context
Response Processing Tips
# Extract URLs only
exa-search "query" | jq -r '.results[].url'
# Get titles and URLs
exa-search "query" | jq '.results[] | {title, url}'
# Extract context string for LLM
exa-search "query" --context | jq -r '.context'
# Check cost
exa-search "query" | jq '.costDollars'
Error Handling
| Error | Solution |
|---|---|
| "EXA_API_KEY not set" | Run fetch-secrets |
| "Unauthorized" | Check API key validity |
| Empty results | Try broader query or different type |
| Rate limited | Wait and retry |
Categories Reference
company- Business websitesresearch paper- Academic papersnews- News articlespdf- PDF documentsgithub- GitHub repositoriestweet- Twitter/X postspersonal site- Personal blogs/sitesfinancial report- SEC filings, reportspeople- People profiles