Web extraction

for AI agents.

Native MCP server for Claude, Cursor, and any AI tool. One API call turns messy pages into clean JSON. Built for agent builders, lead workflows, pricing monitors, and research bots.

Get API Key → See the MCP Setup

100 requests free Paid plans from £19/mo MCP in 30 seconds

MCP Config Python cURL JavaScript

claude_desktop_config.json

// Add to Claude Desktop, Cursor, or any MCP client — done{"mcpServers": {"haunt": {"command": "npx","args": ["@hauntapi/mcp-server"],"env": {"HAUNT_API_KEY": "haunt_your_key" } } } }// Now your AI agent can extract data from any website. // No API calls to write. Just ask in natural language.

One install. Every AI tool.

Available on npm and PyPI. Works with Claude Desktop, Cursor, Windsurf, Cline, and any MCP-compatible client.

npm

npx @hauntapi/mcp-server

PyPI

pip install hauntapi

HTTP API

REST endpoint. Any language. Three lines.

Why Haunt

AI understands pages. MCP connects to your agents. One API call does the rest.

Natural Language Prompts

Describe what you want in plain English. No selectors, no XPath, no fragile parsers that break on layout changes.

Cloudflare Bypass

Automatically handles challenges, CAPTCHAs, and bot detection. Haunt goes where other scrapers can't.

One POST Request

No browser setup. No Selenium. No Puppeteer. Send a URL and a prompt, get structured JSON back in under a second.

Authenticated Scraping

Send cookies and headers to scrape behind login walls. Dashboards, portals, private data — all accessible.

Batch Processing

Extract from up to 10 URLs in a single request. Monitor competitors at scale without rate limit headaches.

Structured Output

Get clean, typed JSON — not raw HTML. Arrays, objects, numbers, strings. Ready for your database or app.

How we compare

Haunt is built for agent workflows first: MCP, browser rendering, structured JSON, and authenticated extraction behind one API.

Feature	Haunt	Firecrawl	Jina Reader
AI-powered extraction	✓	✓	✓
Natural language prompts	✓	✓	—
Hosted MCP endpoint	✓	✓	✓
Cloudflare bypass	✓	✓	Limited
Authenticated scraping	Pro+	Advanced	Cookie forwarding
npm + PyPI packages	✓	✓	✓
Starting price	Free	Free (limited)	Free (limited)
Paid plans from	£19/mo	$19/mo	Usage-based

Real response. No mock data.

This is actual API output. Under a second, clean JSON, zero parsing.

Response — 743ms

// POST /v1/extract // {"url": "https://example.com", "prompt": "Extract the main heading"}{"success": true,"data": {"main_heading": "Example Domain" },"url": "https://example.com","latency\_ms": 743,"tokens_used": 377,"credits_remaining": 99}

Simple pricing

Start free. Scale when you need to. No hidden fees.

£ GBP $ USD € EUR

Free

£0

100 requests / month

• AI-powered extraction
• Natural language prompts
• JSON output
• Cloudflare bypass
• Authenticated scraping
• Priority support Get Free Key

Starter

£19/mo

1,000 requests / month

• AI-powered extraction
• Natural language prompts
• JSON output
• Cloudflare bypass
• Authenticated scraping
• Priority support Start Starter

Pro

£49/mo

5,000 requests / month

• AI-powered extraction
• Natural language prompts
• JSON output
• Cloudflare bypass
• Authenticated scraping
• Priority support Start Pro

Scale

£99/mo

15,000 requests / month

• AI-powered extraction
• Natural language prompts
• JSON output
• Cloudflare bypass
• Authenticated scraping
• Priority support
• Batch processing
• Custom rate limits Start Scale

Overage at £0.02/request. Failed extractions don't count. Cancel anytime.

Built For

AI agent builders who need clean data, not another dependency to manage.

MCP-Powered Agents

Give your Claude or Cursor agent web extraction capabilities. Zero code — just configure and ask.

AI & LLM Pipelines

Feed clean web data into GPT, Claude, or any LLM. Build RAG pipelines, research agents, and chatbots.

Research & Data Science

Collect datasets from the web at scale. Clean, structured data ready for analysis and ML models.

Price Monitoring

Track competitor prices across e-commerce sites. Get clean product data without maintaining fragile selectors.

Lead Generation

Extract company info, contact details, and structured data from business directories and listings.

News Aggregation

Pull articles, headlines, and summaries from news sites. Build custom news feeds and monitoring tools.

FAQ

Quick answers. No fluff.

How is this different from BeautifulSoup?

BeautifulSoup requires CSS selectors that break when sites change. Haunt uses AI to understand page content — you describe what you want in plain English. No more maintaining fragile parsers.

Does it work on Cloudflare-protected sites?

Yes. Haunt automatically detects Cloudflare challenges and routes through our bypass. You don't need to do anything — it just works.

What if extraction fails?

Failed extractions don't count against your quota. You only pay for successful results. Clear error messages tell you what went wrong.

Can I scrape behind a login?

Yes, on the Pro plan. Send your session cookies or auth headers and we'll fetch the page as an authenticated user.

Is there a Python SDK?

Yes! pip install hauntapi. Also available: MCP server for Claude Desktop/Cursor (npx @hauntapi/mcp-server), Postman collection, and it works with any HTTP client in 3 lines of code.

Do you offer refunds?

Start free with 100 requests/month to test it. If you go Pro and aren't happy, contact us within 14 days for a full refund.

Get your API key

Create the key first. If you picked a paid plan, we send you straight to secure Stripe Checkout.

Selected: Free — 100 requests/month, no card.

Get Free API Key

Already have a Haunt key? Choose a paid plan above, paste your key, and go straight to checkout.

Upgrade Existing Key