Firecrawl fetches websites and search results as plain text Claude can work with. Attached as a connector. My agents use it when they research without me.
Four things I use it for
1. Research before the week's content
My content agent fetches sources, articles and numbers before the brief and the carousels are written. If a number is on a slide, it was fetched from a page, not from Claude's memory. That is the difference between a post I can stand behind and one I have to check afterwards.
2. Websites into Claude
A client's site, a competitor's site, the documentation for a tool. If I am at the screen myself, Claude reads it through Chrome. I use Firecrawl when it is many pages at once, or when an agent needs to read them while my browser is busy with something else. Twenty product pages from a competitor come in as plain text in a minute, without me opening a single tab.
3. Search from inside Claude
When I ask about something that has changed since yesterday, prices, features, news, Claude searches through Firecrawl and fetches the pages behind the results. The answer comes with links I can click, and it also works inside an agent running without a browser.
4. Client work: SEO, GEO and leads
The SEO agent fetches competitors' pages and product copy before it writes the client's. The sales agent enriches leads with what is on their website before the draft of the first email is written. Both are Firecrawl in the background.
Whatever has to happen without you needs Firecrawl.
How to connect Claude. Five minutes.
Firecrawl is in Claude's connector directory. You find it under Customize, Connectors, and log in once in the browser, no key to copy. The directory version searches and fetches pages. If you also want to crawl whole sites, you add Firecrawl's own address as a custom connector, and the prompt below does that. Copy it into a Claude project.
You are my research assistant. We are going to connect Claude to Firecrawl so you can read websites and search the web for me. Step 0. Ask about the browser. Start by asking whether you may use my browser (Claude in Chrome), so you can click and fill in the fields for me. If I say yes, you do it and I watch. If I say no, you guide me instead, one step at a time, and wait for my "ok" between each. You never type passwords. I type those myself. Step 1. The account. Open firecrawl.dev and let me create an account. The free plan gives 1,000 credits a month, which is enough to start. Step 2. The connection. In Claude: Customize, then Connectors, the Discover tab. Search for Firecrawl and click Connect to Claude. If I need to crawl whole sites, add it instead with plus, Add custom connector. Name: Firecrawl. URL: https://mcp.firecrawl.dev/v2/mcp-oauth. Add, and approve in the browser. On Claude's free plan there is room for one custom connector. Switch it on in the conversation under the plus, Connectors, and approve the first call with Allow once. Step 3. Check. When I say "connected", fetch the front page of aimedline.dk through Firecrawl and tell me in one sentence what the page is about. If it doesn't work, tell me what is missing instead of guessing. Step 4. My rules. Save them in the project: - When I ask about prices, features, dates or news, you search first. You don't answer from memory. - Every claim that rests on a page comes with the link. - You only fetch pages I ask for, or that a search points to. You don't crawl whole sites without asking. - Keep an eye on usage: one page is one credit, a search is two credits per ten results. Tell me if a task will cost more than 100 credits. Step 5. First task. Ask me what I want researched. Find three sources, summarise each in two lines with a link, and stop there.
Price
Free plan with 1,000 credits a month, no card. One fetched page is one credit, a search is two credits per ten results. Hobby costs 16 dollars a month billed yearly and gives 5,000 credits. Standard costs 83 dollars a month billed yearly, 99 billed monthly, and gives 100,000. For one person doing their own research, free or Hobby goes a long way. Agents fetching many pages a week for several clients end up on Standard.
Through my link you get 10% off your first purchase.
How to get started
1. Create the account through the link at the bottom
Free. No card. The 10% discount is on the link when you later choose a plan.
2. Connect Claude
The prompt above. Five minutes.
3. Give it one task
"Find three sources on X and summarise them with links." That is enough to see what you get: plain text, sources, no tabs.
I am a Firecrawl affiliate. If you buy through my link, I get a share. The price is the same for you, minus the 10%, and I used it in my own systems long before there was a link.
Firecrawl is the part of my setup that reads the web. The rest of the tools, and what each of them is hired to do, are in my tech stack.
