Getting Started with Firecrawl MCP Server

Firecrawl MCP Server is a handy tool if you're looking to scrape and extract data from the web — especially when working with tools that follow the Model Context Protocol (MCP) like Cursor or Claude. This tutorial walks you through the setup and gives you everything you need to get up and running quickly.

What It Does

In short, Firecrawl MCP Server lets you:

• Scrape and crawl websites
• Extract content (text, metadata, links)
• Handle multiple URLs in batches
• Stream results in real time with SSE
• Do all this with retry logic and rate limiting already built-in

If you've got an app, agent, or script that needs structured data from web pages, this tool saves you time.

Prerequisites

• Node.js and npm installed
• A valid Firecrawl API Key (Get one from your Firecrawl account)
• An MCP-compatible environment like Cursor, Windsurf, or your own project

Step-by-Step Setup

1. Basic Configuration

Here's a sample MCP server configuration using Firecrawl:

{
  "mcpServers": {
    "firecrawl-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ],
      "env": {
        "FIRECRAWL_API_KEY": "fc-af1b3ac1a0c2402485402fd0e34da158"
      }
    }
  }
}

Make sure you replace the API key above with your own if you’re not just testing.

You can plug this into your MCP config file, whether you're using Cursor, Claude, or another compatible tool.

Running the Server

To run it directly from the terminal:

FIRECRAWL_API_KEY=fc-yourkeyhere npx -y firecrawl-mcp

This will start the server and expose endpoints that your agent can talk to using the MCP protocol.

Alternatively, if you prefer a global install:

npm install -g firecrawl-mcp
firecrawl-mcp

Use Cases

Here are a few common things people use Firecrawl MCP Server for:

• Scraping product info from ecommerce pages
• Batch extracting text content from news or blog URLs
• Collecting metadata like titles, headers, and links for research
• Building a search layer over unstructured web data

Tips for Better Results

• Use batch mode when possible: It's faster and more efficient.
• Watch rate limits: If you're hitting a lot of URLs, be aware of Firecrawl's rate policies.
• Use SSE mode if you're expecting large outputs or want results streamed live.
• Handle retries: Firecrawl has built-in retry logic, but you should still log and handle errors on your side.

FAQs

Can I run this locally? Yes — works fine both on your machine or in a cloud setup.

Is this only for JavaScript? No — as long as your platform can send MCP requests (usually HTTP-based), it works. Python, Node.js, etc. are all fine.

How do I get an API key? Register at Firecrawl.dev, go to your dashboard, and generate one.

What’s the pricing like? Check their official site — it’s usage-based, with free tiers available.

Final Notes

Firecrawl MCP Server is pretty lightweight but powerful. Once it's set up, your agents or apps can query it just like any other MCP server. Whether you're building a tool for research, competitive analysis, or just need structured web data — this tool can be a big help.

👉 Explore more or grab the config here

Let me know if you want a walkthrough of specific endpoints or example scripts — happy to share.