The MCP scraping service
AI agents run on
Your agent picks one of 50 named tools, we do the scraping, parsed JSON comes back. Works in Claude, Cursor, Windsurf, and any other MCP client.
1,000 free MCP calls. No credit card required.
// in practice
One prompt in. Structured data out.
The model chains tools across sites in one request.
Prompt
// how it works
Three steps.
Zero scraping code.
Setup
Drop in the config
One JSON block in Claude, Cursor, Windsurf or any client speaking streamable HTTP. No SDK, no Node version pinning.
Routing
The model routes by tool name
Your model reads every named tool with full param schemas. We validate inputs server-side before the call runs.
Output
Fields, not HTML soup
Each tool returns a typed schema. Product details, property listings, job postings, search results. No HTML.
// coverage
A complete coverage map
New tools appear in your client's schema as we ship them.
Google, end-to-end
Google SERP, News, Shopping, Images, Events, Trends, Short Videos, AI Overview, AI Mode. Plus Bing.
Maps · Yelp · YellowPages
Google Maps places, reviews, photos, posts. Plus Yelp and YellowPages businesses.
Amazon & Shopify
Amazon ASINs, variants, reviews, sellers, and storefronts. Plus any Shopify catalog.
Zillow · Redfin · Airbnb · Booking
Listings and properties from Zillow and Redfin. Airbnb stays, Booking.com hotels, Google Flights itineraries.
Indeed & Glassdoor
Job listings and individual posting details from Indeed and Glassdoor.
YouTube & Instagram
YouTube search, channels, videos, and transcripts. Plus Instagram public profiles.
Anything else
Any public URL. Markdown by default, or pass a schema for typed JSON. Anti-bot and JS rendering handled server-side.
// clients
Compatible MCP clients
If it speaks streamable HTTP, it speaks HasData.
// install
A six-line diff to your config.
Fifty tools in the schema.
The MCP server is hosted at mcp.hasdata.com/api/mcp. No SDK, no Docker. Paste these lines into Claude Desktop, Cursor, Windsurf, or Claude Code. All tools auto-discover.
// token economy
Save 99% of tokens.
Save your context window.
Raw HTML drowns the model in nav bars, scripts, and ads. HasData parses the page server-side and returns only the fields your agent asked for. One Amazon product page, side by side:
What other scrapers return
11,500 lines of HTML, navigation, scripts, ads. Past Claude's 200k window before the model even starts.
<!doctype html>
<html lang="en"><head><meta charset="utf-8">
<title>Apple iPhone 16 Pro Max, 1TB ...</title>
<script>window.__pre={...}</script>
<style>body{margin:0;...}</style>
</head><body><div id="navbar" class="...">
<a href="/" id="nav-logo">...</a>
<span class="hm-icon-label">All</span>What HasData returns
Structured and schema-validated, ready for the next reasoning step.
// pricing
Flexible Pricing
Start for free, then scale as you grow. One credit balance for MCP and REST.
Free
$0 /moStartup
$49 /moBusiness
$99 /moEnterprise
$249 /mo*Calls vary by tool: web scraping = 1 credit, specialized tools = 5, SERPs = 10.
Explore all the features of our paid plans with a 30-day free trial. No credit card required.
// reliability
99.9% uptime,
engineered every single day.
Everyone claims "99.9% uptime." Behind ours sit synthetic tests, real-time dashboards, proxy health checks, and a self-hosted Kubernetes cluster.
// scale
requests served daily across 50 tools.
p50 latency
1.8s
success rate
99.94%
on site-specific tools
Synthetic tests, every day
We test every API with 10+ parameter variations, several times a day. We validate response shape block by block, not just status code.
Real-time dashboards
Success-vs-failure ratios and p50/p80/p90/p99 latency per tool. Alerts hit our channel with request ID, logs, and cross-service traces.
Proxy health, monitored
Per-retry success rate, traffic per provider, median response size. We rotate out underperforming providers before customers hit failures.
Self-hosted Kubernetes
Dedicated DB and monitoring nodes. Grafana + Prometheus for metrics. ClickHouse for high-volume traces. No managed-cloud markup.
Synthetic test fails
Slack alerted in seconds
Latency confirms
p99 spike on the dashboard
Proxy isolated
success rate drops per provider
Traffic rerouted
fix shipped, loop closes
Grafana + Prometheus · ClickHouse · self-hosted K8s · multi-provider proxy stack.
Read the full reliability post// faq
Frequently asked
questions.
Everything about the MCP server, billing, and which clients work with HasData.
What makes HasData's MCP different from other scraping MCP servers?
Other servers expose a generic scrape function. Your model parses fields from raw HTML or markdown itself. HasData ships each site as a named tool with typed JSON output. Less prompt work, less token cost.
What if the site I need isn't in your 50 tools?
Use the universal web_scraping tool. Pass any URL, get back clean markdown. Or pass a JSON schema for typed fields. For built-in parsing on a specific site, request a custom scraper.
What's a credit and how do I estimate my usage?
One credit covers one basic web scrape without JS rendering. Site-specific tools cost 5 credits per call. SERPs cost 10. So the Free plan's 1,000 credits give you 200 amazon_product calls or 100 SERP queries.
What happens when a target site changes its structure?
Synthetic tests check every tool's response shape daily. When a site changes structure, we patch the parser server-side before you notice. If a field can't be extracted, the tool returns null, not garbage.
Can I customize headers, cookies, geo-targeting, or JS scenarios?
Site-specific tools handle anti-bot, proxies, and fingerprinting automatically. The universal web_scraping tool accepts custom headers, cookies, user-agent, JS rendering, country targeting, and JS scenarios for click and scroll automation.
Can I get more credits or concurrency than the Enterprise plan offers?
Yes. Email support@hasdata.com with your expected volume. We'll put together a custom plan with higher concurrency, larger credit packs, or whatever your workload needs.
How does the MCP server differ from the REST API?
Same infrastructure, two surfaces. REST is for your backend code over HTTP. MCP makes the same 50 tools auto-discoverable inside any MCP-compatible client. Same API key, same credit balance.
Is web scraping with HasData legal?
HasData accesses only publicly viewable content. No login walls, no paywalls, no auth required. Scraping public data is generally legal. Compliance with the target site's terms and your jurisdiction's laws (GDPR, CCPA, EU AI Act) is your responsibility.