You can turn cloud code into a scraping

powerhouse. All you have to do is just

give it access to one MCP server and you

can do things like extracting entire

websites, creating screenshots of that

website, taking all the branding, the

fonts, the colors, the look and feel,

and even pulling structured data for

research. All just using natural

language. If you don't believe me, let

me show you. Now, there are tons of

methods and MCP servers and tools you

can use for scraping. I just happen to

use firecrawl and you'll see it in a

bunch of my videos in the past. And the

reason for that is all you have to do is

give it a simple input which is

typically a URL or a bulk set of URLs

and you can choose what kind of data you

want back. So if I go to my agency

website promptadvisor.com

and I take this and we paste it in here,

we can just click on start scraping and

you'll see on the next screen you have

tons of options you can focus on. Yes,

you can just scrape the website as is,

but this isn't new. But what you can do

is you can click on things like

branding, for example, click on start

scraping, and then while this goes

through the URL, it'll come back not

just with the scrape of the website, but

also all the information about the

fonts, the colors, all the look and feel

elements that make my prompt advisor's

website actually native to me. So just

as a small example, you could see this

is the branding. This is my logo, my

assets, my landing page, all the primary

colors, fonts, and typography. And this

is just one small sliver of what you can

do. Now, my goal is to bring this power

into Cloud Code. So you can just say

something like go and search these

following websites. Go and grab the

branding from this one. Go and grab

screenshots from the bottom and the top

of the page of XYZ website. So if you go

to the docs and you go to the MCP server

section on the right hand side, you can

see right here there's different

sections for running in cloud code as

well as nen and other platforms. So if I

click on cloud code right here, all you

have to do is copy paste this command as

is and you're going to want to swap this

with your API key. You can get this API

key by just signing in. They have a

pretty generous free tier of 500

requests. So, if you go into the

dashboard, you'll see at the right hand

side right here, this is your API key.

I'm going to just take this and attach

it to that request and I'll send it over

in cloud code. So, all you have to do is

just enter that command as is with your

API key and then within seconds it

should say done and then it won't show

up or actually function within this same

chat or session. You're going to have to

respin up a brand new session. So,

you'll see right here it says firecrawl

is there. So it exists but it status

failed. But if I open a brand new

session and do slashmcp

load it up. Now we can see that this is

functional. So what I'll do is just to

show you that it does work. I'll do

shift tab and I'll go into yolo mode and

I'll say something like can you tell me

all the tools we have access to with the

fire crawl MCP? Maybe give me some use

cases as to what I can use each tool

for.

So we'll send this over and we'll get

back something like this that says we

have firecrawl scrape which is good for

scraping content from different URLs

parsing PDF documents even firecrawl map

firecrawl search which is what we just

used a few minutes ago and then the

crawl function which can go on multiple

portions and subpages of the same

website crawl status extract which uses

LLM capabilities to extract content on a

particular page so you can search for

different components or elements of said

page and this is the quick guide. Now,

in order to be able to create a

claude.md file, we'll make this into an

asset in the codebase since right now

it's empty. So, claude MD won't really

do anything for us. So, I'll say, okay,

great. Can you put a full understanding

of all the different functionalities of

this MCP server into a file that we call

firecrawl mcp server guide domarkdown.

So, then this will be sent and then

we'll be able to send it over and double

check from there exactly how to move

forward. And now that the guide is set

up, we can click on this right here. You

can see it's pretty comprehensive. And

going through there's a table of

contents and it breaks down every single

parameter that Claude should be aware

of. And once I type in slashinit, it

will initialize a brand new claude MD

file which won't be as detailed, but it

will have the TLDDR cheat sheet of

exactly how to use this resource. Now

that we have this set up, I'll go

through a mental model of how to use

this API just as a quick cheat sheet so

that we can open a brand new session and

experiment with just using natural

language like I promised to do things

like take screenshots, extract content,

and even more importantly extract the

look and feel and design of a particular

website. So, while there's some sub

functionality that we can use, it's good

to use this as your core cheat sheet.

So, you can scrape, screenshot, extract,

crawl, search, and map. In terms of what

to scrape, you could do a client brand

audit where you can scrape an entire

website and see is the copywriting up to

snuff when it comes to the brand

strategy and the tone of voice. If

you're a designer or a web developer,

then the screenshot functionality might

be handy to create portfolio documents

where it screenshots your entire web

page as one continuous document. You can

use the extract functionality to go on

any website that has tons of listings.

That could be e-commerce, it could be a

job platform, it could be Indeed, what

have you. And then crawling as well as

searching can both be used for research

and competitor analysis. And map is

really important to look at the site map

of an entire website. So if there are

multiple components and many subdomains,

it's a really good way to be able to

look at the entire topography of a

particular domain. So theory aside,

let's actually put this to the test. So

all I'm going to do is go on YOLO mode.

It'll go on shift tab tab tab and then

we'll take the URL of my agency website

and we'll just send a natural language

command. So I'll say the following. Use

the firecrawl MCP server to do the

following on this particular website.

And I'll just paste this in and I'll

just paste the URL. I'll go back. So, I

want to be able to extract all the

branding of this website as well as take

screenshots of the entire website from

top to bottom so I can see it in one

unified image. And then I also want to

scrape and understand what this company

does. And I want to possibly crawl and

see what other web pages there are as

well. Maybe screenshot those and put all

of these together in one folder that are

neatly put together. So, we could send

this over, right? And then paste it and

we'll see what it comes back with. So

you can see it's invoking the MCP

multiple times in this case to scrape.

You can even take a little bit of a

peak. So this is the website. These are

the parameters. It's using the

functionality type screenshot. And then

if we close this up, you can see all the

links from the map. So this is the map

function. It's mapping out the services,

the about us, everything involved.

Pretty straightforward. And then you can

see it's already setting up this folder

on the lefth hand side. And while this

is running, it's important to note that

you can use this on the free tier, the

hobby, the starter, and your only

changes are the number of requests that

you can send, as well as the number of

concurrent requests, which in plain

English means how many different calls

can you make to the same API, the crawl

API, at the same time. The one good

thing about the MCP server is if it

can't run more than the limit, so let's

say on the hobby plan, it's five, it

will cue and wait until it's ready to

go, then it'll resend the request. So,

now it's halfway and you can see right

here on the to-do list, we're still

creating the branding guide document. If

you take a little peek at where we're

at, it takes all the color palette of my

website, my typography, all the fonts,

the spacing, the layout, the logo,

everything you could need. And how could

you use this? Let's say you build RFPs

or requests for proposals or tenders or

anything that you build at scale and you

want to tailor it to either your

company, the company you work for, or a

company you're reaching out to. These

kinds of MCP server tools are amazing

because you can literally apply them to

any documents at scale. And if you've

watched my prior video on using clawed

code to create document files, you can

now marry both worlds where you can take

the file itself and then change the

branding and the look and feel using

something like this far crawl MCP. And

by the way, because I'm such a super

user of this platform, they gave me a

special link in the description below

that if you use that will give you way

more credits than the standard free

tier. So now that it's finished, it's

the moment of truth. So we have the

branding guide. We already saw that the

content. So this is all the company

overview of what my agency does. This is

mwah. It goes through pretty much every

part of my website. And if we close

that, the part that's probably

interesting to you are the screenshots.

So I can go on something like the

homepage. And you can see right here, it

takes a beautiful end to end top to

bottom picture of my website as well as

the other components of my website. And

this is just running one singular

command. What would happen if you had

multiple agents with different tasks?

One crawling, one scraping and doing

research, one taking screenshots, and

based off of that screenshot, basically

executing a task or doing analysis based

on whatever input is received. I'm not a

big fan of using many MCP servers in

general because it basically bloat your

context window, but this is one of the

few MCPs that I use on a recurring

basis. And that's pretty much it. So, I

just wanted to walk you through how to

use this MCP server and all the

associated tools. And you can include

this in whatever systems you want. For

example, I use this in my own AI

strategy system that I put together for

my exclusive community. And it's been so

helpful that people are actually closing

deals. They're doing audits and

strategies just using things like Fry

Crawl combined with some other things to

be able to look at a website, tailor the

branding, and apply it in a variety of

scenarios. So, if you found this

helpful, then please let me know down in

the comments below. helps the video,

helps the channel, and I'll see you in

the next