Day 2 Building a MCP · AI Automation Society

Day 2 Building a MCP

Just set up Firecrawl MCP and scraped a real site without writing a single line of code 🔥

Spent the morning connecting Firecrawl to Claude Code and honestly, it's a game-changer for anyone doing data extraction work.

Here's what I did:

Connected Firecrawl's MCP server to Claude Code (literally just pasted an install command)
Added my API key to .env and reloaded
Tested it on a simple site — got clean markdown back instantly
Created a Firecrawl cheat sheet (Claude wrote it for me) explaining each tool and when to use it
Went for a real use case — found a job board and asked Claude to extract structured data into CSV
Used Plan Mode first — let Claude map out the approach before executing

The result: A clean CSV with hundreds of job listings, extracted and formatted automatically. No web scraping libraries, no parsing logic, no manual API wrangling.

What blew my mind:

You literally just describe what you want extracted and Claude handles it
Firecrawl's smart enough to handle pagination and messy HTML
The structured data output is actually usable (not garbage text)
Switching to Plan Mode first saved me from mistakes — Claude thought through the approach before executing

The setup was smooth: