web-research

Web Research with x402 APIs

Access Exa (neural search) and Firecrawl (web scraping) through x402-protected endpoints.

Setup

See rules/getting-started.md for installation and wallet setup.

Quick Reference

When to Use What

See rules/when-to-use.md for detailed guidance.

Exa Neural Search

Semantic search that understands meaning, not just keywords:

npx agentcash@latest fetch https://stableenrich.dev/api/exa/search -m POST -b '{
  "query": "startups building AI agents for customer support",
  "numResults": 10,
  "type": "neural"
}'

Options:

• query - Search query (required)
• numResults - Number of results (default: 10, max: 25)
• type - "neural" (semantic) or "keyword" (traditional)
• includeDomains - Only search these domains
• excludeDomains - Skip these domains
• startPublishedDate / endPublishedDate - Date range filter
• category - Filter by content type: "company", "research paper", "news", "pdf", "github", "tweet", "personal site", "linkedin profile", "financial report"
• Tip: Use category: "linkedin profile" for people/profile discovery

Returns: List of URLs with titles, snippets, and relevance scores.

Find Similar Pages

Find pages semantically similar to a reference URL:

npx agentcash@latest fetch https://stableenrich.dev/api/exa/find-similar -m POST -b '{
  "url": "https://example.com/article-i-like",
  "numResults": 10
}'

Great for:

• Finding competitor products
• Discovering related content
• Expanding research sources

Extract Text Content

Get clean, structured text from URLs:

npx agentcash@latest fetch https://stableenrich.dev/api/exa/contents -m POST -b '{
  "urls": [
    "https://example.com/article1",
    "https://example.com/article2"
  ]
}'

Options:

• urls - Array of URLs to extract
• text - Include full text (default: true)
• highlights - Include key highlights

Cheapest option ($0.002) when you already have URLs and just need the content.

Direct Answers

Get factual answers to questions:

npx agentcash@latest fetch https://stableenrich.dev/api/exa/answer -m POST -b '{"query": "What is the population of Tokyo?"}'

Returns a direct answer with source citations. Best for:

• Factual questions
• Quick lookups
• Verification of claims

Firecrawl Scrape

Scrape a single page to clean markdown:

npx agentcash@latest fetch https://stableenrich.dev/api/firecrawl/scrape -m POST -b '{"url": "https://example.com/page-to-scrape"}'

Options:

• url - Page to scrape (required)
• formats - Output formats: ["markdown", "html", "links"]
• onlyMainContent - Skip nav/footer/ads (default: true)
• waitFor - Wait ms for JS to render

Advantages over WebFetch:

• Handles JavaScript-rendered content
• Bypasses common blocking
• Extracts main content only
• LLM-optimized markdown output

Firecrawl Search

Web search with automatic scraping of results:

npx agentcash@latest fetch https://stableenrich.dev/api/firecrawl/search -m POST -b '{
  "query": "best practices for react server components",
  "limit": 5
}'

Options:

• query - Search query (required)
• limit - Number of results (default: 5)
• scrapeOptions - Options passed to scraper

Returns search results with full scraped content for each.

Cloudflare Website Crawl

Crawl multiple pages from a website with browser rendering. Async two-step pattern.

Step 1: Start the crawl (paid, $0.10)

npx agentcash@latest fetch https://stableenrich.dev/api/cloudflare/crawl -m POST -b '{
  "url": "https://example.com",
  "limit": 10,
  "depth": 1,
  "formats": ["markdown"]
}'

Returns 202 with {"token": "jwt..."}.

Step 2: Poll for results (SIWX, free)

npx agentcash@latest fetch "https://stableenrich.dev/api/cloudflare/jobs?token=JWT_TOKEN"

Poll every 3-5 seconds until complete.

Parameters:

• url (required) — starting URL
• limit (default 10, max 25) — max pages
• depth (default 1, max 3) — max link depth
• formats — ["markdown", "html", "json"]
• render (default false) — execute JavaScript
• options.includePatterns / excludePatterns — URL wildcards

Good for: crawling docs sites, scraping multiple pages, building sitemaps.

Workflows

Deep Research

• [ ] (Optional) Check balance: npx agentcash@latest balance
• [ ] Search broadly with Exa
• [ ] Find related sources with find-similar
• [ ] Extract content from top sources
• [ ] Synthesize findings

npx agentcash@latest fetch https://stableenrich.dev/api/exa/search -m POST -b '{"query": "AI agents in healthcare 2024", "numResults": 15}'

npx agentcash@latest fetch https://stableenrich.dev/api/exa/find-similar -m POST -b '{"url": "https://best-article-found.com"}'

npx agentcash@latest fetch https://stableenrich.dev/api/exa/contents -m POST -b '{"urls": ["url1", "url2", "url3"]}'

Blocked Site Scraping

• [ ] Try WebFetch first (free)
• [ ] If blocked/empty, use Firecrawl with waitFor for JS-heavy sites

npx agentcash@latest fetch https://stableenrich.dev/api/firecrawl/scrape -m POST -b '{"url": "https://blocked-site.com/article", "waitFor": 3000}'

Cost Optimization

• Use Exa contents ($0.002) when you already have URLs
• Use WebSearch/WebFetch first (free) and fall back to x402 endpoints
• Batch URL extraction - pass multiple URLs to Exa contents
• Limit results - request only as many as needed