MCP server

The Model Context Protocol (MCP) is the open standard AI assistants use to call external tools. ScrapeIt runs a hosted MCP server: point Claude, Cursor, VS Code or any other MCP client at one URL, give it an API key, and the assistant can scrape pages, search Google, Google Maps, Amazon, Booking.com, Yelp, Walmart, eBay, Zillow, Apartments.com, Allegro, Redfin, Tripadvisor, Airbnb and Yellow Pages, start crawls and batches and check your credit balance - no code, no local install.

Setting	Value
URL	https://api.scrapeit.io/api/v1/mcp
Transport	Streamable HTTP, stateless - MCP revision 2026-07-28 and the 2025 revisions
Auth	x-api-key: YOUR_SCRAPEIT_API_KEY or Authorization: Bearer YOUR_SCRAPEIT_API_KEY

Every tool call runs the same engine as the REST API: the same proxies, anti-bot handling, SSRF protection, concurrency limits and billing outcome rules.

Connect your client

Create an API key in Workspace → API Keys with only the scopes the assistant needs, then add the server to your client.

Claude Code

claude mcp add --transport http scrapeit https://api.scrapeit.io/api/v1/mcp \
  --header "x-api-key: YOUR_SCRAPEIT_API_KEY"

Cursor

Add the server to ~/.cursor/mcp.json (all projects) or .cursor/mcp.json (one project):

{
  "mcpServers": {
    "scrapeit": {
      "url": "https://api.scrapeit.io/api/v1/mcp",
      "headers": { "x-api-key": "YOUR_SCRAPEIT_API_KEY" }
    }
  }
}

VS Code (GitHub Copilot agent mode)

Add .vscode/mcp.json to the workspace. VS Code asks for the key once and stores it securely:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "scrapeit-api-key",
      "description": "ScrapeIt API key",
      "password": true
    }
  ],
  "servers": {
    "scrapeit": {
      "type": "http",
      "url": "https://api.scrapeit.io/api/v1/mcp",
      "headers": { "x-api-key": "${input:scrapeit-api-key}" }
    }
  }
}

Claude Desktop

Claude Desktop reaches remote servers with custom headers through the mcp-remote bridge (needs Node.js). Add this to claude_desktop_config.json (Settings → Developer → Edit Config) and restart Claude:

{
  "mcpServers": {
    "scrapeit": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.scrapeit.io/api/v1/mcp",
        "--header",
        "x-api-key:${SCRAPEIT_API_KEY}"
      ],
      "env": { "SCRAPEIT_API_KEY": "YOUR_SCRAPEIT_API_KEY" }
    }
  }
}

Any other MCP client

Use the Streamable HTTP transport with the URL above and send the key in x-api-key or as Authorization: Bearer. The server keeps no session: every request is authenticated on its own, so GET and DELETE on the endpoint answer 405.

Tools and scopes

Each tool needs the same API key scope as its REST endpoint. A key without it gets a PERMISSION_DENIED error naming the missing scope in details.required_scope; the other tools keep working.

Tool	What it does	Scope	Same as
scrape_url	Scrape one page; returns markdown by default, plus extractRules / aiExtractRules results.	scrape:web	POST /api/v1/scrape/web
google_search	One Google results page as JSON (organic, ads, People also ask, related, knowledge panel).	scrape:web	GET /api/v1/scrape/google/search
google_maps_search	One page of Google Maps places (twenty per page) as JSON.	scrape:web	GET /api/v1/scrape/google/maps/search
google_maps_place	One place by place_id, data_id, cid or url: hours, attributes, photos, top reviews.	scrape:web	GET /api/v1/scrape/google/maps/place
google_maps_reviews	Reviews of one place by sort and keyword query, up to 5 per call: rating, text, translation, likes, photos, visit details, owner response.	scrape:web	GET /api/v1/scrape/google/maps/reviews
amazon_search	One Amazon search results page as JSON.	scrape:web	GET /api/v1/scrape/amazon/search
amazon_product	One Amazon product by ASIN or URL as JSON.	scrape:web	GET /api/v1/scrape/amazon/product
amazon_best_sellers	One Amazon Best Sellers page (50 ranked products) by category or url.	scrape:web	GET /api/v1/scrape/amazon/best-sellers
booking_search	Booking.com properties for a destination and dates (destination, checkin, checkout).	scrape:web	GET /api/v1/scrape/booking/search
booking_hotel	One Booking.com property by url or hotel_id; room prices with dates.	scrape:web	GET /api/v1/scrape/booking/hotel
yelp_search	One page of Yelp businesses for find_loc (and optionally find_desc).	scrape:web	GET /api/v1/scrape/yelp/search
yelp_business	One Yelp business by alias or url.	scrape:web	GET /api/v1/scrape/yelp/business
walmart_search	One walmart.com search results page as JSON.	scrape:web	GET /api/v1/scrape/walmart/search
walmart_product	One walmart.com product by item_id or url.	scrape:web	GET /api/v1/scrape/walmart/product
ebay_search	One www.ebay.com search page (60, 120 or 240 listings) with filters as JSON.	scrape:web	GET /api/v1/scrape/ebay/search
ebay_item	One www.ebay.com item by item_id or url, optionally with the full description.	scrape:web	GET /api/v1/scrape/ebay/item
zillow_listing	zillow.com homes for sale, for rent or sold in a US city and state, with filters.	scrape:web	GET /api/v1/scrape/zillow/listing
zillow_property	One zillow.com property by url: facts, schools, price and tax history, listing agent.	scrape:web	GET /api/v1/scrape/zillow/property
apartments_listing	apartments.com rentals in a US city and state, with filters.	scrape:web	GET /api/v1/scrape/apartments/listing
apartments_property	One apartments.com property by url: floor plans, units, fees, scores.	scrape:web	GET /api/v1/scrape/apartments/property
allegro_listing	allegro.pl offers by phrase, category_id, url or product_id, with filters and page_token paging.	scrape:web	GET /api/v1/scrape/allegro/listing
allegro_offer	One allegro.pl offer by offer_id or url.	scrape:web	GET /api/v1/scrape/allegro/offer
redfin_search	Homes of a redfin.com area by location or url, with filters; the first 50 unless limit (up to 350).	scrape:web	GET /api/v1/scrape/redfin/search
redfin_property	One redfin.com property by property_id or url: facts, Redfin Estimate, records, tax, history.	scrape:web	GET /api/v1/scrape/redfin/property
tripadvisor_search	One tripadvisor.com hotels or restaurants list (30 places) by location, geo_id or url; hotel prices for checkin/checkout.	scrape:web	GET /api/v1/scrape/tripadvisor/search
tripadvisor_place	One tripadvisor.com hotel or restaurant by url.	scrape:web	GET /api/v1/scrape/tripadvisor/place
airbnb_search	One page of airbnb.com stays (18 listings) for a location, dates and guests.	scrape:web	GET /api/v1/scrape/airbnb/search
airbnb_listing	One airbnb.com listing by listing_id or url, priced for your dates.	scrape:web	GET /api/v1/scrape/airbnb/listing
yellowpages_search	One page of yellowpages.com businesses (30) for q and a US location.	scrape:web	GET /api/v1/scrape/yellowpages/search
yellowpages_business	One yellowpages.com business by url: contacts, hours, categories, reviews.	scrape:web	GET /api/v1/scrape/yellowpages/business
glassdoor_listing	One page of www.glassdoor.com jobs (30) for q and a location, with salary, date and type filters; cursor pages on.	scrape:web	GET /api/v1/scrape/glassdoor/listing
glassdoor_job	One www.glassdoor.com job by job_id or url: description, pay, job type, skills and the employer profile.	scrape:web	GET /api/v1/scrape/glassdoor/job
glassdoor_companies	One page of www.glassdoor.com companies (10) by location, sector, industry and rating.	scrape:web	GET /api/v1/scrape/glassdoor/companies
glassdoor_company	One www.glassdoor.com company by company_id or url: ratings, counts, offices.	scrape:web	GET /api/v1/scrape/glassdoor/company
crawl_start	Start a Website Crawler job.	scrapers:run	POST /api/v1/scrapers/crawler/jobs
crawl_status	Status and progress of a crawl.	scrapers:read	GET /api/v1/scrapers/jobs/{id}
batch_start	Start a batch over many URLs.	scrape:batch	POST /api/v1/scrape/batch/web
job_status	Status, progress and cost of any job (batch, crawl or single request).	scrapers:read	GET /api/v1/jobs/{id}
job_stop	Stop a running batch or crawl.	scrapers:stop	DELETE /api/v1/scrape/jobs/{id}
get_usage	Credit balance, plan limits and usage of the current period.	usage:read	GET /api/v1/usage/me

The arguments are the ones of the REST endpoint (for example url, jsRendering, outputFormat, extractRules, aiExtractRules, proxyType, proxyCountry and screenshot for scrape_url) and are validated by the same rules; your client shows each tool's full argument list.

Pricing

A tool call costs exactly what the REST request costs - see Request cost and the product pages for the price lists. crawl_status, job_status, job_stop and get_usage are free. Tool calls share your plan's concurrency slots and request rate limit with your REST traffic.

Responses and limits

• Page content. scrape_url returns a short JSON summary (status, cost, files, content.json, ai_extraction) followed by the page itself as plain text - markdown unless you ask for another outputFormat.
• Time. A tool call that scrapes answers within 90 seconds, like the REST site APIs: a step that would end later is not started, and the call answers SCRAPE_TIMEOUT without charging it. A larger timeoutMs on scrape_url is cut to what is left of the 90 seconds.
• Size. A tool result carries at most 100,000 characters (about 25,000 tokens), so one page cannot flood the assistant's context. Longer content is cut and ends with a [Scrapeit: truncated, …] note; ask for specific fields with extractRules or aiExtractRules instead of whole pages.
• Request size. One MCP request body is limited to 1 MiB. Send large URL lists through the REST Batch Scraping API instead of batch_start.
• One message per request. JSON-RPC batches (an array of messages in one body, dropped from MCP in revision 2025-06-18) are refused with HTTP 400 Invalid Request; send each tool call as its own request. Every request counts against the request rate limit.
• No change notifications. The tool list does not change while a client is connected, so the server advertises listChanged: false: a subscriptions/listen request is acknowledged with no subscriptions and closed at once, and GET (a 2025-era notification stream) answers 405.
• Results in Files. A site API tool call keeps its JSON result in Files, and scrape_url its page and content, as the same REST request does.
• Results of async jobs. crawl_start and batch_start return the job at once; their results are in the dashboard, Files and your webhooks.

Errors

A failed call comes back as a tool result marked as an error, carrying the same body as the REST API, for example:

{
  "error": {
    "code": "INSUFFICIENT_CREDITS",
    "message": "Insufficient credits.",
    "correlation_id": "…"
  }
}

Code	Meaning
PERMISSION_DENIED	The key lacks the tool's scope (details.required_scope).
VALIDATION_ERROR	An argument failed the endpoint's validation; the message names it.
INSUFFICIENT_CREDITS	The balance cannot cover the estimated cost. Nothing ran.
CONCURRENCY_LIMIT_EXCEEDED	All concurrent slots of the plan are busy. Retry when a request finishes.
INTERNAL_ERROR	A platform failure. Quote the correlation_id to support.

A missing or revoked key is refused before any tool runs with HTTP 401 INVALID_API_KEY, and too many requests with HTTP 429 RATE_LIMIT_EXCEEDED. All codes are listed in Error codes.

Security

• One key per client, least scopes. A research assistant that only reads job status needs scrapers:read and usage:read; leave scrape:web, scrapers:run and scrape:batch off keys that should never spend credits.
• Scraped pages are untrusted text. A page can contain instructions aimed at the assistant. Keep your client's per-tool approval on for the paid tools, and watch the Usage page.
• Your workspace only. A key acts inside the workspace it was created in; no tool reaches another workspace or any admin function. The key is never returned by a tool or written to logs.

Related pages

• Authentication & API keys - creating keys and choosing scopes.
• Web Scraping API - every scrape_url option in detail.
• Billing & credits - when a request is charged.