Paste one config block that runs npx @crawlbase/mcp into your tool's MCP config, add your Crawlbase token, and restart. The web tools appear in your agent in about a minute.

Product / Web MCP Server

Web MCP Server.
Make your LLMs web-aware in seconds.

Connect Claude, Cursor, Windsurf and any MCP-compatible agent to live web data with one config.
Your agents read the now, not a training cutoff.

Integrate free Explore on GitHub

Claude · Cursor · Windsurf · n8nPowered by the Crawling APIMCP always free

Live agent tool calls1.24M req/minStreaming

200npmjs.com/package/reactCA213ms

200techcrunch.com/category/artificial-intelligenceAU177ms

200wikipedia.org/wiki/TransformerGB116ms

200google.com/search?q=mcp+serversNL188ms

200wikipedia.org/wiki/TransformerJP161ms

200reddit.com/r/MachineLearningUS105ms

200news.ycombinator.com/newestBR211ms

200amazon.com/dp/B08N5WRWNWGB134ms

301pypi.org/project/langchainNL146ms

200docs.stripe.com/api/chargesNL60ms

200docs.stripe.com/api/chargesBR204ms

200github.com/trendingNL130ms

200amazon.com/dp/B08N5WRWNWDE214ms

200google.com/search?q=mcp+serversUS120ms

404google.com/search?q=mcp+serversSG89ms

200weather.com/weather/todayDE99ms

404news.ycombinator.com/newestUS128ms

200stackoverflow.com/questions/79100123CA87ms

200techcrunch.com/category/artificial-intelligenceAU122ms

200pypi.org/project/langchainIN59ms

200producthunt.com/posts/notionUS53ms

200docs.stripe.com/api/chargesNL139ms

200github.com/trendingNL155ms

200news.ycombinator.com/newestUS120ms

200techcrunch.com/category/artificial-intelligenceBR144ms

200techcrunch.com/category/artificial-intelligenceUS102ms

200npmjs.com/package/reactCA213ms

200techcrunch.com/category/artificial-intelligenceAU177ms

200wikipedia.org/wiki/TransformerGB116ms

200google.com/search?q=mcp+serversNL188ms

200wikipedia.org/wiki/TransformerJP161ms

200reddit.com/r/MachineLearningUS105ms

200news.ycombinator.com/newestBR211ms

200amazon.com/dp/B08N5WRWNWGB134ms

301pypi.org/project/langchainNL146ms

200docs.stripe.com/api/chargesNL60ms

200docs.stripe.com/api/chargesBR204ms

200github.com/trendingNL130ms

200amazon.com/dp/B08N5WRWNWDE214ms

200google.com/search?q=mcp+serversUS120ms

404google.com/search?q=mcp+serversSG89ms

200weather.com/weather/todayDE99ms

404news.ycombinator.com/newestUS128ms

200stackoverflow.com/questions/79100123CA87ms

200techcrunch.com/category/artificial-intelligenceAU122ms

200pypi.org/project/langchainIN59ms

200producthunt.com/posts/notionUS53ms

200docs.stripe.com/api/chargesNL139ms

200github.com/trendingNL155ms

200news.ycombinator.com/newestUS120ms

200techcrunch.com/category/artificial-intelligenceBR144ms

200techcrunch.com/category/artificial-intelligenceUS102ms

01 Live demo

One config in. Live web out.

The Web MCP Server, typed live. Paste the config, then watch an agent pull clean markdown from a live page. Hover to pause and read.

ready

keys 1-2 switch · click to pauserun your own URL

Run your first request in minutes. Up to 10,000 free requests, no credit card.Start free

02 Capabilities

Live web data, wired into your agent.

Everything an agent needs to read the live web, behind one config and three tools.

setup

One-config setup

Paste a single config block that runs npx @crawlbase/mcp, add your token, and the web tools appear in about a minute.

clients

Works where you work

Claude Desktop, Cursor, Windsurf and n8n out of the box, plus any other agent that speaks the Model Context Protocol.

fresh

Grounded in the now

No more stale context or hallucinated facts. Tools return live, rendered pages, so answers reflect what is on the web today.

tools

Three web tools

crawl for live HTML, crawl_markdown for clean markdown, and crawl_screenshot for a full-page image.

engine

Powered by the Crawling API

Every tool call runs on the Crawling API, so JavaScript rendering, residential proxies and anti-bot handling are built in.

free

MCP always free

The server itself is free. You only pay for the Crawling API requests it makes, and up to 10,000 are free.

03 How it works

A one-minute setup.

Pick a tool, paste a config, add your token, and start calling live web tools.

Select your tool

Claude Desktop, Cursor, Windsurf or n8n. Any MCP-compatible client works.

Add the Web MCP Server

Paste one config block into your tool's MCP config file. No build, no service to run.

Add your token

Drop your Crawlbase token into the config and restart. Up to 10,000 requests are free.

Call the tools

Ask your agent to crawl, return markdown or screenshot any URL, live.

04 Use cases

What agents do with a live web.

USE / 01Research

Answers grounded in now

Let a research agent read live pages instead of guessing from a training cutoff.

USE / 02Coding

Current docs in your IDE

Pull up-to-date docs and references straight into Cursor or Windsurf as you build.

USE / 03Workflows

Live data in n8n

Wire crawl tools into n8n automations so flows act on what the web says today.

USE / 04RAG

Fresh retrieval context

Feed clean markdown and rendered pages into retrieval, kept current on every call.

USE / 05Monitoring

Agents that watch the web

Have an agent check prices, listings or status live, then act on the result.

USE / 06Any client

One server, any agent

Connect any MCP-compatible tool to the same server with the same config.

05 Pricing

Add the sites you crawl, see the price.

Add the sites you crawl with their monthly volume and request type. We group them by difficulty and type, then price each group on its combined volume, so the more you crawl, the cheaper it gets.

100k / mo

No sites yet. Add one above to start your estimate.

Estimated monthly cost

$0/ mo

≈ $0.00 blended per 1,000 requests

Up to 10,000 requests free. No credit card.

Start free Crawling over 1B a month? Talk to us →

06 Notes

Good to know.

MCP is free

The server itself costs nothing. You only pay for the Crawling API requests your tools make, and up to 10,000 are free with no credit card.

Open source

The server is open source on GitHub. Run it with npx or self-host it for n8n and remote setups.

Fully documented

Setup for each client and every tool is covered in the MCP docs, with copy-paste configs.

GDPR and CCPA compliant

Crawlbase applies consumer-protection standards globally, with fairness and transparency built into how data is handled.

07 Why Crawlbase

The live web, built for agents.

The Web MCP Server runs on the same network that serves 70,000+ developers, with the Crawling API underneath. No proxies to buy, no browsers to run, nothing to patch when a site changes.

1 min

Setup, one config block

Tools: crawl, markdown, shot

70K+

Developers on the network

Free

Web MCP Server, up to 10,000 free requests

One config, any MCP agent, and a residential network underneath every tool call.

08 FAQ

MCP questions.

A Model Context Protocol server that connects your AI agent to live web data. With one config block it gives Claude, Cursor, Windsurf and other MCP-compatible agents tools to crawl pages, return clean markdown and capture screenshots, all grounded in the present.

Any MCP-compatible client. It is tested with Claude Desktop, Cursor, Windsurf and n8n, and works with other agents that speak the Model Context Protocol.

Three: crawl for live rendered HTML, crawl_markdown for clean markdown extracted from a URL, and crawl_screenshot for a full-page screenshot.

Yes. The Web MCP Server itself is free. You only pay for the Crawling API requests it makes, and up to 10,000 requests are free with no credit card.

Paste one config block that runs npx @crawlbase/mcp into your tool config, add your Crawlbase token, and restart. The web tools appear in about a minute.

Yes. The Web MCP Server is powered by the Crawling API, so JavaScript rendering, residential proxies and anti-bot handling all apply to every tool call.

Give your agents the live web.
One config, one minute.

Free to begin, with up to 10,000 free Crawling API requests. MCP stays free.