@webpeel
基本信息
Web data API that just works. Pass any URL, get clean markdown — Cloudflare, JavaScript, CAPTCHAs handled automatically. No config needed. 97.6% success rate across 500+ domains. Free tier included.
使用下面的配置,将此服务器添加到你的 MCP 客户端。
{
"mcpServers": {
"webpeel": {
"command": "npx",
"args": [
"-y",
"webpeel",
"mcp"
]
}
}
}Your complete web toolkit — fetch, search, screenshot, extract, monitor, and interact with any website. Handles JS rendering, Cloudflare, CAPTCHAs, and 55+ domain extractors automatically. 65-98% token savings. Describe what you want in plain English. Examples: 'read https://stripe.com/pricing', 'screenshot bbc.com on mobile', 'search for best AI frameworks 2024', 'extract product prices from amazon.com/dp/...', 'watch stripe.com/pricing for price changes', 'get YouTube transcript from youtu.be/...'. For JavaScript-heavy SPAs (React, Vue, Next.js, Polymarket, Airbnb, etc.), mention 'render' or 'use browser' in your task. For infinite scroll or lazy-loaded content, say 'scroll to bottom'. For bot-protected sites (Cloudflare), say 'stealth mode'. If you get sparse or empty content, retry and mention 'render' — the site likely requires JavaScript.
Fetch any URL and return clean, LLM-optimized markdown. 65-98% fewer tokens than raw HTML. Automatically handles: web pages, YouTube transcripts (with timestamps), PDFs, JS-rendered SPAs, Cloudflare-protected sites, and 55+ domain-specific extractors (Amazon, Reddit, GitHub, etc.). IMPORTANT: Use render=true for ANY JavaScript-heavy site (React, Vue, Angular, Svelte, Next.js, SPAs). Known SPAs that need render=true: Polymarket, Airbnb, Booking.com, Expedia, Indeed, Zillow, Google, and more. If content is sparse or empty, ALWAYS retry with render=true before concluding the page has no content. Use actions= to interact with the page before extraction (scroll, click, type, wait). Use stealth=true for Cloudflare-protected or bot-blocked sites (auto-enables render). Use question= for instant Q&A (no LLM needed). Use summary=true for a short summary. Use budget=N to distill to N tokens.
Capture a screenshot of any web page. Returns the page as an image for visual inspection. Supports mobile, tablet, and desktop viewports. Use mode='design' for AI-powered design analysis and suggestions. Use mode='compare' with compare_url to diff two pages visually. Use full_page=true to capture the entire scrollable page.
Search the web or discover all pages on a site. Pass query= to search the web and get ranked results with titles, URLs, and snippets. Pass url= to map/crawl a domain and discover all its pages. Use depth='deep' for multi-source research that synthesizes answers from multiple pages. Smart search detects intent for restaurants, products, flights, hotels, and more.
Extract structured JSON data from any URL. No LLM needed for built-in schemas. Pass fields=['price','title','description'] to extract specific named fields. Pass schema={...} with a full JSON schema for custom structured output. Built-in schemas: product, article, recipe, job, event, contact, business, review, listing. Works on Amazon, Yelp, LinkedIn, job boards, e-commerce sites, and any web page.
Track a web page for content changes over time. Call once to take a snapshot, call again to get a diff. Use selector= to monitor a specific CSS element (e.g. a price, a status badge). Use webhook= for persistent monitoring with automatic notifications when content changes. Use interval= to set how frequently to check ('1h', '30m', '1d'). Ideal for: price tracking, job listing changes, release monitoring, competitor updates.
Automate interactions with any web page using a real browser. Click buttons, fill forms, select dropdowns, scroll, wait for elements, and press keys. Returns extracted content and optionally a screenshot after all actions complete. Use for: logging into sites, submitting forms, navigating multi-step flows, interacting with dynamic content that requires user input. Actions: click, type, fill, scroll, wait, press, hover, select.
Webpeel is a zero-config MCP server that fetches any URL and converts it into clean markdown or structured JSON. It automatically handles Cloudflare, bot detection, and JavaScript, giving AI agents high-quality, token-efficient content. Designed for developers who need reliable web scraping without Docker, Redis, or complex setup.
Install and run with a single command: npx webpeel "url". The server exposes seven MCP tools: webpeel (fetch as markdown), webpeel_read (reader mode), webpeel_see (screenshots with device emulation), webpeel_find (search + crawl), webpeel_extract (structured data with JSON schemas), webpeel_monitor (track page changes), and webpeel_act (click, type, scroll, interact). Free tier includes 2,000 fetches per month with no credit card required.
Webpeel requires zero configuration (no Docker, no Redis), achieves a 97.6% success rate on difficult sites, and reduces token usage by 65–98% through clean markdown output. It also includes 55+ built-in extractors for structured data.
None. Run `npx webpeel "url"` directly – no Docker, Redis, or local setup needed.
The README does not specify data persistence or residency. Data is returned in real time to the requesting MCP client.
The free tier allows 2,000 fetches per month. No details on paid tiers or rate limits beyond that are provided.
The server uses the standard MCP protocol. Authentication is not mentioned; the free tier requires no credit card.
Official Porkbun MCP server, maintained by Porkbun. Check domain availability and pricing, register, renew and transfer domains, and manage DNS records, nameservers, URL forwarding, glue records, DNSSEC and domain contac
Remote MCP server for Mythex, an AI app builder. Claude, Cursor or Codex get a cloud sandbox to build full-stack apps: files and bash, live preview, Postgres database, file storage, secrets, checkpoints, browser tests an
Launch your product from your AI agent. Connect Claude Code, Codex, Cursor or ChatGPT to Launchdog over MCP and your agent drafts the listing, picks a launch week and books it. 50 launches a week, ranked only by human vo
Query your team's drift, vulnerability, and migration data from any AI assistant. Vibgrate MCP connects Cursor, Claude, ChatGPT, Windsurf, or VS Code to Vibgrate Cloud: 51 tools for DriftScores, CVEs and EOL runtimes, up
Give an agent a place to put large output. Instead of returning 40k characters of a build log into the conversation, it posts the file to DocBin and hands back one short URL. Docs render with syntax highlighting and mark
评论