Do I need Firecrawl for this course?

What Firecrawl actually is

Firecrawl is a tool that pulls clean, readable content out of web pages. Feed it a URL, and instead of the raw HTML with all the menus, ads, pop-ups, and sidebars, you get the actual text of the article in a format that's easy for a large language model to work with. If you've ever used a "reader mode" in a browser, the idea is similar: strip the page down to the part you came for.

In Claude Code, Firecrawl shows up as an MCP server — an external tool that Claude can call when it needs to extract a web page. When Claude decides to fetch and read a URL, it routes the request through Firecrawl, gets back the clean text, and writes it to a file in your project folder. That's why Joe's demo produces organized markdown files in a firecrawl folder: the tool is extracting each source page and saving it as it goes.

Why Joe uses it in the Module 1 demo

In the Module 1 video, Joe runs the same research prompt — "find everything you can about the Center for Cooperative Media and give me a report" — in Claude Code and Claude on the web at the same time. Claude on the web finishes a polished summary faster. Claude Code is slower because it asks for permissions and, among other things, uses Firecrawl to pull each source page into a local file before summarizing.

That slower path is the whole point of the demo. The web app's summary lives only in that chat. Claude Code leaves behind a folder of extracted source material that any future session can read back. Firecrawl is what makes that folder of extracted sources cleaner and more usable than what you'd get from raw HTTP fetches.

If you run the same research prompt without Firecrawl installed, Claude Code will still do the research and still save whatever it reads into files. The files just won't be as clean, and the overall workflow won't match the video quite as closely. That's the whole difference.

You don't have to install it

The course does not require Firecrawl at any point. Nothing in Modules 1 through 4 or the final project assumes you have it. Claude Code can already read web pages without it — the output is just a little less clean, and the research demo in Joe's Module 1 video will look slightly different from yours. That's the whole difference.

If you're short on time or energy this week, skip Firecrawl entirely. Come back to it later only if you hit a specific task where you actually want the cleaner content extraction it offers.

If you want it anyway, ask your agent

If you do decide you want Firecrawl set up, the way you install it in this course is the same way you install anything else: you ask your CLI agent to walk you through it. Inside an active Claude Code, Gemini CLI, or Codex CLI session, paste something like:

I want to install the Firecrawl MCP server so you can use it
to extract clean content from web pages. Walk me through it
step by step. Tell me what to click, what to copy, and where
to paste it. Handle any config files yourself. I'm new to
this, so assume I'm starting from zero.

The agent will hand you instructions one step at a time. Some of those steps might involve visiting firecrawl.dev in your browser, creating an account, copying an API key, and pasting it back into the conversation when the agent asks for it. That's fine — you're still following the rule. The rule is that you don't type install commands into a separate terminal yourself. Clicking around on a website or pasting a key into a chat message doesn't violate anything.

This "ask your agent to walk you through it" pattern is the default way to install any new tool, MCP server, or plugin in this course. If you ever find yourself about to type something into a terminal that isn't cd or your agent's launch command (claude, gemini, codex), stop and go ask your agent instead.

What to do if your demo doesn't match Joe's

One participant in the forums ran the same research-the-person prompt Joe demoed and got a terminal response but no files. She thought Firecrawl was the reason. It wasn't — the fix was adding one sentence to her prompt about where to save the output.

If your research session isn't producing files, start with Why didn't Claude Code save any files? before blaming Firecrawl's absence. It's almost always a prompting issue, not a tooling issue.