Web scraping in Claude Code,
in one command
Claude Code picks from 57 named tools, we handle the proxies and the parsing, and the result lands in your repo as a changed file. Nothing to install.
1,000 free credits. No credit card required.
You ask. Claude Code scrapes the rest.
Each prompt names a file in the repo. The answer is a change to it.
Prompt
Agent trace
Prompt
Agent trace
Prompt
Agent trace
Prompt
Agent trace
Prompt
Agent trace
Command
Agent trace
Commit .mcp.json and the setup travels with the repo. A teammate pulls, approves the server once, and has the same tools.
Add the MCP server. That's the whole setup.
MCP, the default
already done if you ran the command aboveNo binary, no download. The model reads 57 tool schemas and routes by name. A project-scoped server waits for your approval, so run claude once to approve it:
The Claude Code plugin, for the jobs you rerun often enough to deserve a command
Same key, same credits, no separate signup. Default to MCP. Add the plugin when a prompt hardens into a routine.
The usual suspects, already parsed.
The whole results page
Google SERP, Light SERP, AI Mode, AI Overview, News, Shopping, Images, Trends. Plus Bing.
Businesses, ready to call
Google Maps places, reviews, photos and posts. Plus Yelp and YellowPages.
Prices, current and before
Amazon products and sellers. Any Shopify catalog.
Homes, stays and fares
Zillow, Redfin, Airbnb, Booking, Google Hotels, Google Flights.
Channels, posts and transcripts
YouTube search, video, channel and transcript. TikTok profile, posts and keyword search. Instagram.
Openings and employers
Indeed for the openings, Glassdoor for the employers behind them.
Anything else
Any public URL. Markdown for reading, or name the fields and get JSON. Proxies, rendering and captchas stay on our side.
1,000 free credits is a finished job.
Free
Startup
Business
RecommendedGrowth
Monthly credit volume
Custom price based on required volume
Past 100M credits a month, or terms the self-serve plans do not cover. We shape the contract around your workload.
How do I scrape websites in Claude Code?
Add the HasData MCP server with one command: claude mcp add hasdata -s project -t http https://mcp.hasdata.com/api/mcp. Claude Code discovers 57 scraping tools (SERP, Maps, Amazon, any public URL), calls them by name, and writes the results into files in your repo. Proxies, rendering and parsing run on HasData’s side.
Isn't WebFetch enough?
Often, yes. It reads one page at a known URL, and it's free. It doesn't return parsed search results, doesn't pass anti-bot protection, and isn't built for two hundred rows. If your job is a lookup, keep WebFetch. If it's a list, that's what the 57 tools are for.
What's a credit and how far does 1,000 go?
Tool calls run 1–15 credits. A plain page read is 1, Light SERP and Maps are 5, a full SERP is 10. The free 1,000 is exactly one 200-row enrichment run, or 1,000 page reads.
Where does my API key live?
As a variable. With single quotes in the add command, ${HASDATA_API_KEY} lands in .mcp.json unexpanded, so the file commits safely and the key stays in your environment. Double quotes would let the shell expand it into the file. Don’t.
What does -s project change?
It writes .mcp.json into the repository instead of your local config, so the server definition gets reviewed, versioned and pulled like any other file. Without the flag the scope is local, your machine only.
How do long jobs behave on the free tier?
A page load caps at 300 seconds, and a request that fails there is not charged. Free-tier concurrency is 1, so a loop over a file runs sequentially. Paid plans start at 15 concurrent requests and go up.
What if the site I need isn't covered?
The universal web_scraping tool takes any public URL and returns markdown, raw HTML, or JSON in a schema you define. For built-in parsing on a specific site, request a custom scraper.
MCP server or the plugin?
MCP is the default. Claude reads the tool schemas and combines them inside any prompt you write. The plugin ships 13 fixed skills behind slash commands and runs them through the hasdata CLI. When a prompt hardens into a routine you type every week, that routine is a plugin command. Same key and the same credit balance either way.
How does MCP differ from the REST API?
Same infrastructure, two surfaces. REST is for your backend code over HTTP. MCP makes the same 57 tools auto-discoverable inside Claude Code. Same API key, same credit balance.
Is this legal?
HasData accesses only publicly viewable content: no login walls, no paywalls. Scraping public data is generally legal. Compliance with the target site’s terms and your jurisdiction’s laws (GDPR, CCPA) is your responsibility.
All integrations · The MCP server itself, including full platform coverage
The web, as tool calls.
Web scraping in Claude Code, minus the infrastructure. No headless browsers, no proxy pools, no parsers to maintain, just structured data back.
One command · 1,000 free credits · no card