Web pages to Markdown and link lists with Firecrawl

Web pages to Markdown and link lists with Firecrawl

An n8n workflow that turns a list of web addresses into Markdown. For each address it asks Firecrawl for the page as Markdown, plus
the page's title, description and links. It works through the list in batches and pauses between them to respect the rate limit.

What is inside

• workflow.json: the workflow as n8n exports it, 17 nodes, seven of them sticky notes. It holds no credential. One note is a long
  overview with setup steps, and the others are short hints about limits and the parts you must fill in.
• README.md: these notes.

What we changed from the original copy

• The HTTP Request node was linked to a credential in its author's n8n. The link is gone, so select your own.
• Retrieve Page Markdown and Links called Firecrawl's version 1 address, which Firecrawl switches off on 10 December 2026. It now
  calls https://api.firecrawl.dev/v2/scrape. The request body is the same.
• Markdown data and Links now also returns the address of the page in a field called url, so each row says which page it came
  from. The links field turned the list into one text with a stray space at the end. It now holds one link per line.
• The Wait node had only an amount. Its unit is now written out as seconds.
• One of the two example addresses was the original author's own site. It is now https://example.com/.
• The long sticky note named its author and a website. It was rewritten with the same setup steps, and the note beside the HTTP
  Request node now says the key goes after the word Bearer.
• Nothing else.

Before you start

• An n8n instance, on n8n Cloud or on your own server.
• A Firecrawl account and its API key. The free plan needs no card.
• A list of web addresses, one per row in a column named Page, or the two examples in the file.
• A place to put the results. The workflow ends in an empty node where you connect your own.

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. In n8n, add a Header Auth credential. Set its name to Authorization and its value to the word Bearer, a space and your API key.
   Select it in Retrieve Page Markdown and Links.
3. In Example fields from data source, replace the two addresses with your own. To read them from a sheet or a database, connect
   that node to Get urls from own data source and give it a Page column.
4. Connect your own storage after Connect to your own data source. A Google Sheets, Airtable or database node works.
5. Press Execute workflow and check the rows in your storage.

How a run goes

1. When clicking ‘Test workflow’ starts the run, and Get urls from own data source is an empty node where your own list comes in.
2. Example fields from data source holds a list named Page, and Split out page URLs makes one item per address.
3. 40 items at a time keeps the first 40 items. 10 at a time takes ten of them.
4. Retrieve Page Markdown and Links calls Firecrawl once for each of the ten, asking for the Markdown and the links.
5. Markdown data and Links keeps the page address, title, description, Markdown content and links of each reply.
6. Connect to your own data source is where you store the rows. Wait then pauses 45 seconds and 10 at a time takes the next ten.
7. When no addresses are left, the run ends.

Change the look

• Which pages are read: the list in Example fields from data source, or your own source.
• How many are read in one run: the Max Items value of 40 items at a time.
• How many go to Firecrawl at once and how long it waits: the batch size of 10 at a time and the amount in Wait.
• What is kept from each page: the fields of Markdown data and Links.
• Where the rows go: the node after Connect to your own data source.

Good to know

• The old address: Firecrawl's docs, read on 12 November 2026, say the version 1 scrape, crawl, search, map and account endpoints stop
  on 10 December 2026 and answer with HTTP 410 after that. Mindket's copy uses version 2, which takes a Bearer key in the
  Authorization header.
• Limits: Firecrawl's docs list the free plan at 1,000 credits a month, 2 concurrent requests and 10 scrape requests a minute. The
  Hobby plan is $16 a month billed yearly, or $19 billed monthly, for 5,000 credits and 100 scrape requests a minute. The
  batch of 10 and the pause of 45 seconds fit the free plan. A plain scrape costs one credit.
• Only 40 at a time: 40 items at a time drops every address after the first 40. Raise the number if your server has the memory,
  or run the list in pieces.
• One bad page: a page that Firecrawl cannot fetch makes the HTTP node fail, and that stops the run. Pages already passed on stay
  in your storage.
• Nothing is stored: the workflow ends in an empty node. Until you connect storage, the rows exist only in the run.
• Check a site's terms before you read it in bulk.
• Your data: the addresses go to Firecrawl, which fetches the pages.
• Not run: nothing here was run against Firecrawl. These notes come from the file, n8n's source and Firecrawl's docs, read on 12
  November 2026.