Local Browser MCP Server
WebHarvy's Local Browser MCP Server connects to Claude Desktop, Codex (ChatGPT) and LM Studio in one click. Just describe what you want to scrape in plain English - the AI drives WebHarvy's real browser for you. Unlike browser-only MCP tools, WebHarvy loads Cloudflare-protected pages, handles login-gated content and navigates JavaScript-heavy sites that simpler scrapers can't reach.
Pay Once, Scrape Forever
MCP Server Features
Whether you're building AI-powered data pipelines or just want to ask Claude or ChatGPT to pull data from a website, WebHarvy puts a full-featured scraper at your AI assistant's fingertips - with no monthly subscription and no cloud API keys to manage.
- 01
Claude, ChatGPT & LM Studio
Connect Claude Desktop, Codex (ChatGPT) or LM Studio from WebHarvy's Settings with a single click - WebHarvy adds itself to the client's configuration for you. You can connect more than one client at the same time.
- 02
Scrape Using Plain English
Just tell your AI assistant what you want: "scrape all product names and prices from this page." The AI translates your instructions into scraping actions - no selectors, no configuration needed.
- 03
No Monthly Fees. Runs Locally.
Unlike cloud-based MCP scraping APIs that charge monthly, WebHarvy is a one-time purchase. It runs entirely on your machine with no API keys to manage. Pair it with a local model in LM Studio and nothing leaves your computer.
- 04
Load Any Website - Including Cloudflare
WebHarvy uses a full Windows browser, so it loads Cloudflare-protected pages, login-gated content and JavaScript-heavy SPAs that browser-only MCP tools or simple HTTP scrapers cannot handle.
- 05
24 Browser Actions
Your AI assistant can navigate, click, type, take screenshots, follow pagination, load infinite-scroll lists and use WebHarvy's pattern detection to find repeating lists and tables from a single sample item.
- 06
You Stay in Control
Watch the AI work in WebHarvy's visible browser and step in at any time. When a CAPTCHA or bot check appears, the assistant asks you to solve it and then carries on.
What Your AI Assistant Can Do
WebHarvy gives your AI assistant 24 browser actions. It uses the same browser you see in the WebHarvy window, so it works with pages you have already opened or logged in to.
- 6
Navigation
Open URLs, go back and forward, reload, go to the next page of a paginated listing (WebHarvy finds the "Next" link or page number automatically) and open iframes.
- 8
Reading the Page
Read page text, list links, buttons and inputs, find a repeating list or table from one sample item using WebHarvy's pattern detection, extract matching elements, get HTML, take screenshots and wait for content to appear.
- 7
Interacting
Click elements by selector or visible text, click every matching element (for example all "Read more" links), hover, type and submit, pick dropdown options, press keys and run JavaScript.
- 3
Scrolling & Loading More
Scroll the page or a panel inside it, and load every item of an infinite-scroll or "Load more" list.
Who Uses WebHarvy's MCP Server?
AI Developers & Prompt Engineers
Build data pipelines that feed LLM workflows with fresh web data, triggered directly from Claude, ChatGPT or a local model in LM Studio.
Researchers & Analysts
Let your AI assistant gather structured data from websites while you focus on analysis - describe what you need in plain language and get a spreadsheet back.
Automation Engineers
Wire WebHarvy into broader agentic workflows through MCP, combining AI reasoning with reliable, browser-based web scraping.
Business Intelligence Teams
Use AI assistants to orchestrate competitive research, price monitoring and market data collection without writing scraping scripts.
No-Code Power Users
Already using Claude Desktop, ChatGPT or LM Studio for daily work? Add web scraping to your AI toolkit without any programming knowledge.
MCP & AI Scraping Guides
- MCP ServerScrape from AI Assistants via MCPStep-by-step setup guide for connecting WebHarvy to Claude Desktop, Codex (ChatGPT) or LM Studio using the MCP server.
- AI ScrapingWeb Scraping Using AI with WebHarvyOverview of all AI-powered scraping modes in WebHarvy - LLM extraction, MCP server and more.
- AI ScrapingAI Web Scraping - Pros and ConsWhen does AI-driven scraping make sense, and when should you stick with traditional point-and-click configuration?
Frequently Asked Questions
What is an MCP server and how does WebHarvy use it?
MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude and ChatGPT communicate with external tools. When WebHarvy runs as an MCP server, Claude Desktop, Codex or LM Studio can send it instructions to browse websites and extract data - all on your local machine.
Which AI tools can connect to WebHarvy's MCP server?
WebHarvy connects to Claude Desktop, Codex (ChatGPT) and LM Studio with one click from Settings. Other MCP clients that can start a local (stdio) MCP server should also work using the same command, though they haven't been tested yet.
Do I need coding skills to use the MCP server?
No coding is required. Once WebHarvy is connected to your AI assistant, you interact with it using plain English. The AI handles the technical side - you just describe what data you want.
Does my data get sent to the cloud when using MCP?
WebHarvy's MCP server runs locally on your Windows machine and doesn't open any network port, so other computers can't reach it. The page content WebHarvy reads is sent to your AI assistant, so with Claude or ChatGPT it goes to that provider for processing. Use a local model in LM Studio if you want everything to stay on your computer.
Can it scrape Cloudflare-protected and login-gated pages?
Yes - this is a key advantage over browser-only MCP tools like Chrome MCP. WebHarvy runs a full Windows browser that loads pages exactly as a real user would, passing Cloudflare checks and other bot-detection systems. It can also scrape pages that require a login, since you can browse to the logged-in state before scraping begins.
What happens when a CAPTCHA appears?
WebHarvy detects visible reCAPTCHA, hCaptcha, Cloudflare challenges and "Access Denied" pages and tells the AI assistant. The assistant then asks you to solve it in the WebHarvy window and tries again once you confirm.
Can it handle pagination and infinite scroll?
Yes. The assistant can go to the next page of a listing (WebHarvy finds the "Next" link or page number automatically) and load every item of an infinite-scroll or "Load more" list.
Can the MCP server scrape JavaScript-heavy websites?
Yes. WebHarvy uses a real browser engine, so it handles JavaScript rendering, dynamic content and single-page applications - the same sites that trip up simpler HTTP-based scrapers.
Can I still use WebHarvy manually without the MCP connection?
Absolutely. The MCP server is an optional mode. You can continue using WebHarvy's visual point-and-click interface independently, or switch between manual and AI-driven scraping as needed.
What formats can I export scraped data to?
Scraped data can be exported to Excel, CSV, JSON, XML or TSV files, or saved directly to a MySQL, SQL Server or Oracle database.
Is there a monthly subscription fee?
No. WebHarvy is a one-time purchase - pay once and use it for as long as you like, with no recurring charges.
Try WebHarvy free for 15 days
Download the free trial and connect WebHarvy to Claude Desktop, Codex or LM Studio in minutes. No monthly subscription - pay once, scrape forever.