Web Research
Route paid Exa and Firecrawl research through agentcash when free WebSearch or WebFetch are not enough for semantic or blocked pages.
Overview
Web Research is an agent skill most often used in Idea (also Validate, Build) that routes paid Exa and Firecrawl calls via agentcash when semantic or scrape-heavy research beats free WebSearch.
Install
npx skills add https://github.com/merit-systems/agentcash-skills --skill web-researchWhat is this skill?
- Decision tree: WebSearch vs Exa vs WebFetch vs Firecrawl scrape by query type and page behavior
- agentcash CLI setup: install, balance, redeem invite, Base/Solana USDC funding
- Exa neural search (~$0.01), answer (~$0.01), find-similar (~$0.01), contents (~$0.002)
- Firecrawl scrape (~$0.0126) when JS-heavy or blocked pages break free fetch
- Troubleshooting table for CLI not found, insufficient balance, and transient payment failures
- Exa search, answer, find-similar at $0.01
- Firecrawl scrape at $0.0126
- Exa contents at $0.002
Adoption & trust: 1.4k installs on skills.sh; 11 GitHub stars; 1/3 security scanners passed (skills.sh audits).
What problem does it solve?
Your agent keeps using free search and fetch on conceptual queries or blocked sites and returns shallow or empty results.
Who is it for?
Agent builders who already use agentcash USDC and need a clear when-to-pay playbook for research during discovery and build.
Skip if: Simple keyword lookups where free WebSearch suffices, or teams that cannot use paid network APIs or crypto-funded wallets.
When should I use this skill?
Need web information and the decision tree points to semantic Exa search, paid scrape, find-similar, contents, or answer—not free keyword search or WebFetch.
What do I get? / Deliverables
The agent picks the right paid Exa or Firecrawl action with funded agentcash wallet and returns ranked semantic results or full page text for downstream specs.
- Research results from chosen Exa or Firecrawl endpoint
- Documented tool choice per query in agent trace
Recommended Skills
Journey fit
Spans multiple journey phases - primary shelf plus alternate fits below.
Idea/research is the first canonical shelf because builders most often install this skill to explore markets and concepts before building. Research subphase fits the decision tree that picks semantic Exa search, scraping, find-similar, contents, and quick answer APIs over free tools.
Where it fits
Run Exa semantic search on a vague opportunity space before you commit to a product thesis.
Scrape competitor pricing pages with Firecrawl when WebFetch returns empty on JS-rendered sites.
Pull Exa contents from several API doc URLs into one context block for integration planning.
Use find-similar on a high-ranking article to map content gaps for your distribution plan.
How it compares
Paid research orchestration skill—not a replacement for free WebSearch; use when semantic depth or scraping justifies micro-payments.
Common Questions / FAQ
Who is web-research for?
Solo and indie builders running Claude Code, Cursor, or similar agents with agentcash who want governed paid web research.
When should I use web-research?
In Idea/research for market and concept discovery; in Validate for competitor and landing proof; in Build when docs or APIs need scraped or semantic sources beyond free tools.
Is web-research safe to install?
It requires network spend and wallet access—review Security Audits on this page, fund only what you need, and audit CLI commands before automated runs.
SKILL.md
READMESKILL.md - Web Research
# Getting Started ## Setup 1. **Install the agentcash CLI:** ```bash npm install -g agentcash ``` 2. **Check wallet:** ```bash npx agentcash@latest balance ``` 3. **Fund wallet** (if needed): - Redeem invite: `npx agentcash@latest redeem YOUR_CODE` - Or run `npx agentcash@latest accounts` to get Base or Solana deposit links and wallet addresses ## Troubleshooting | Issue | Solution | |-------|----------| | "Command not found" | Run `npm install -g agentcash` | | "Insufficient balance" | Fund wallet with USDC | | "Payment failed" | Check balance, retry (transient errors) | # When to Use x402 Web Research ## Decision Tree ``` Need web info? | +-- Simple keyword search? --> Use WebSearch (free) | +-- Semantic/conceptual search? --> Use Exa search ($0.01) | +-- Need to scrape a specific URL? | | | +-- WebFetch works? --> Use WebFetch (free) | | | +-- Blocked/JS-heavy? --> Use Firecrawl scrape ($0.0126) | +-- Find similar pages? --> Use Exa find-similar ($0.01) | +-- Already have URLs, need text? --> Use Exa contents ($0.002) | +-- Quick factual answer? --> Use Exa answer ($0.01) ``` ## Exa vs WebSearch | Aspect | WebSearch | Exa Search | |--------|-----------|------------| | Price | Free | $0.01 | | Search type | Keyword | Neural/semantic | | Best for | Quick lookups | Deep research | | Results | Web snippets | Ranked with meaning | | Use when | Simple queries | Complex concepts | **Use Exa when:** - Query is conceptual ("companies solving X problem") - Need semantic similarity - Researching a topic in depth - Standard search isn't finding what you need **Use WebSearch when:** - Query is straightforward keywords - Quick fact check - Don't need deep results - Budget is a concern ## Firecrawl vs WebFetch | Aspect | WebFetch | Firecrawl | |--------|----------|-----------| | Price | Free | $0.0126 | | JS rendering | No | Yes | | Blocked sites | Often blocked | Bypasses many | | Output | Raw HTML | Clean markdown | | Main content | No filtering | Extracts main content | **Use Firecrawl when:** - WebFetch returns blocked/403/empty - Site uses heavy JavaScript - Need clean markdown, not HTML - Content is behind soft paywall **Use WebFetch when:** - Site works with simple fetch - HTML format is fine - Cost is a concern - Testing before x402 call ## Exa Search vs Firecrawl Search | Aspect | Exa Search | Firecrawl Search | |--------|------------|------------------| | Price | $0.01 | $0.0252 | | Output | URLs + snippets | URLs + full content | | Best for | Finding sources | Search + immediate read | **Use Exa search** to find relevant URLs, then selectively extract content. **Use Firecrawl search** when you want search results AND their full content in one call. ## Cost Comparison For a typical research task (find 10 sources, read 3): | Approach | Cost | |----------|------| | WebSearch + WebFetch | $0 (if sites allow) | | Exa search + contents | $0.01 + $0.006 = $0.016 | | Firecrawl search (5) | $0.0252 | | Exa search + Firecrawl scrape (3) | $0.01 + $0.0378 = $0.0478 | ## Recommendation 1. **Start free**: Try WebSearch + WebFetch first 2. **Fall back to x402**: Use Exa/Firecrawl when free tools fail 3. **Exa for discovery**: Use Exa search to find URLs 4. **Cheapest extraction**: Use Exa contents ($0.002/call) for known URLs 5. **Firecrawl for tough sites**: Use when JS rendering or bypass needed --- name: web-research description: | Neural web search and content extraction using x402-protected APIs. Better than WebSearch for deep research and WebFetch for blocked sites. USE FOR: - Deep web research and investigation - Finding similar pages to a reference URL - Extracting clean text from web pages - Scraping sites that block standard fetchers - Getting direct answers to factual questions - Research requiring multiple sources - Crawling multiple pages from a website TRIGGERS: - "research", "investigate", "deep dive", "find sources" - "similar to", "pages l