How to Use Firecrawl with n8n for Web Automation

Bex TuychievBex Tuychiev
Nov 10, 2026 (updated)

TL;DR

  • Firecrawl ships its own verified n8n node: it is published as @mendable/n8n-nodes-firecrawl, maintained by Firecrawl, and covers Firecrawl /search, /scrape, /interact, /crawl, /map and /agent.
  • n8n Cloud connects in one click: Connect to Firecrawl creates an n8n-linked Firecrawl team with 10,000 free Firecrawl credits. The earlier launch offer with a Hobby plan and 100,000 credits has expired.
  • Self-hosted n8n uses the same node: an instance owner or admin installs it once from the nodes panel, and every member can use it with a Firecrawl API key.
  • Firecrawl /agent runs inside the node: Extract Data With AI Agent waits up to 300 seconds by default (600 at most), and the async operation plus Get Agent Status handles longer jobs.
  • A v1 Base URL breaks Agent and Interact: credentials from older node versions can point at https://api.firecrawl.dev/v1, so set them to https://api.firecrawl.dev/v2.
  • The HTTP Request node still has a job: it reaches API parameters the node doesn't expose and imports cURL commands straight from the Firecrawl API reference.

Why use Firecrawl with n8n?

n8n is an open-source workflow automation platform that connects 2,000+ apps and services, and in October 2025 it raised a $180 million Series C at a $2.5 billion valuation. Most business workflows built on it need fresh web data at some point. That is where hand-built scrapers tend to break: sites change their HTML, JavaScript-heavy pages don't render, and someone has to keep the scraping infrastructure running.

Firecrawl handles the web data side of this integration. The API turns pages into clean Markdown or structured JSON and takes care of JavaScript rendering, so each n8n node downstream gets data it can use. With Firecrawl connected, your n8n workflows can:

  • Search the web: run a Firecrawl /search query and pull content from the results in one call
  • Scrape pages: turn any URL into Markdown or structured JSON with Firecrawl /scrape
  • Interact with pages: click buttons, fill forms and read data that only appears after an action with Firecrawl /interact
  • Crawl sites: gather pages across a whole website with Firecrawl /crawl
  • Map sites: list every URL on a domain with Firecrawl /map
  • Research with an AI agent: describe the data you need and let Firecrawl /agent find the sources

Ready for More? Check out our collection of n8n web scraping workflow templates for production-ready examples. For visual AI development, explore our LangFlow tutorial.

This tutorial shows how to set up the Firecrawl node on n8n Cloud and on self-hosted n8n, when to reach for the HTTP Request node instead, and how to build a chatbot that can access live web data. By the end, you'll have a working example that you can adapt for competitor monitoring, content aggregation, or Firecrawl /agent enrichment.

This guide assumes you have a basic understanding of Firecrawl's API. If you're new to Firecrawl, spend a few minutes with the official documentation first.

Getting started with n8n

n8n's own tutorials are a good place to start. Their first workflow tutorial and courses cover the canvas, nodes and credentials you'll use in this guide.

Let's explore the basics through a real example: a workflow that connects WhatsApp and Slack for bidirectional messaging. You can explore this workflow interactively to see how the pieces fit together.

n8n workflows consist of nodes connected by edges that pass data between operations. The first part of this workflow converts Slack messages into WhatsApp messages and sends them to your specified contact.

The workflow begins with a Slack trigger node that listens for events like new messages in channels, support tickets, or mentions.

n8n provides various trigger types to start workflows: manual triggers, scheduled timers, cron jobs, and application-specific triggers like this Slack example.

After the trigger fires, two nodes retrieve the channel ID where the message was posted. A conditional node then determines whether the message contains only text or includes media files.

For media messages, the workflow downloads the content before sending it to WhatsApp. Text-only messages go directly to the WhatsApp node.

The reverse direction, converting WhatsApp messages to Slack, has more steps because WhatsApp messages often include audio and file attachments.

Notice the "download media" node in the workflow. It uses n8n's HTTP Request node, which can send GET, POST, PUT, and DELETE requests to any API. You'll use the same node later for Firecrawl API calls the Firecrawl node doesn't expose.

You can also notice red warning flags next to the Slack and WhatsApp nodes. These indicate authentication errors: since these nodes connect to external accounts, you need to set up credentials first.

To configure authentication, select a node and press Enter to open its settings, then click "Create new credential." You can also configure other node settings here, such as filtering by channel ID.

Try the quickstart examples or the templates library before continuing with this tutorial to get hands-on experience with n8n's interface.

Two ways of using Firecrawl with n8n

Now that you understand workflow basics and node types, let's connect Firecrawl to n8n.

Firecrawl is available in n8n as a verified node (@mendable/n8n-nodes-firecrawl), built and maintained by Firecrawl and verified by n8n. The same node runs on n8n Cloud and on self-hosted n8n. What changes between the two is how you install it and connect your Firecrawl account.

On the node, you pick a Resource and then an Operation. The resources are Scraping, Crawling, Map & Search, Agent, Interact, Browser, Account, and Extract (Legacy), with 31 operations between them in version 2.2.0. The Firecrawl n8n docs describe every operation, and the source lives in the n8n-nodes-firecrawl repository.

Method 1: n8n Cloud with the Firecrawl node

On n8n Cloud, the Firecrawl node is the fastest way to start: one-click connect, no credential juggling. Open the nodes panel, search for "Firecrawl", install the node, and click Connect to Firecrawl. n8n creates a Firecrawl team linked to your n8n account and adds 10,000 free Firecrawl credits. The launch offer that included a free Hobby plan and 100,000 promotional credits has expired. To see those credits on the Firecrawl dashboard, switch to the n8n-linked team in the team selector.

When to use this method: Choose n8n Cloud when you want to start immediately without infrastructure setup.

We'll build a practical example: a GitHub trending scraper that finds repositories in your chosen field every morning.

Setting up the workflow

Start by signing up for n8n Cloud and creating a workflow from your dashboard.

Add the Schedule Trigger:

  1. Click the plus icon and search for "Schedule Trigger"
  2. Configure it to run daily at 5 AM
  3. This node will automatically start your workflow each morning

Add the Firecrawl node:

  1. Search for "Firecrawl" and add the node to your canvas
  2. Connect it to the Schedule Trigger node
  3. Set Resource to Agent and Operation to Extract Data With AI Agent

Configuring the Firecrawl /agent call

The Firecrawl /agent operation runs open-ended research from a single prompt and returns structured results.

Configure your request:

  1. Prompt: "Find 5 trending MLOps repositories on GitHub this week and return name, description, stars and URL for each"
  2. Schema Type (optional): leave it at None, or set a schema if you want the same fields back on every run
  3. Max Wait Time: leave the default of 300 seconds, or raise it to 600 for bigger research tasks

Firecrawl node in n8n with Resource set to Agent and Operation set to Extract Data With AI Agent

The Agent resource with Extract Data With AI Agent selected. Max Wait Time defaults to 300 seconds.

Click "Execute step" to test it. If no MLOps repositories are trending, try a broader topic like "Python" to check that the setup works. You should see the repository data in the output panel, and the node is ready to run on your schedule.

If a job takes longer than the Max Wait Time, the node returns a timeout status with the jobId, even though the job may still finish on the Firecrawl side. Pass that ID to a second Firecrawl node with Get Agent Status, or switch to Start AI Agent Job (Async) for jobs that can run past 10 minutes.

When should you use the HTTP Request node instead?

Use n8n's HTTP Request node when you need an API parameter the Firecrawl node doesn't show, or when you want to paste a request from the Firecrawl API reference without mapping fields by hand. The HTTP Request node has an Import cURL button that fills in the method, URL, headers and body for you.

This is the same /agent request as a cURL command:

curl -s -X POST "https://api.firecrawl.dev/v2/agent" \
  -H "Authorization: Bearer $FIRECRAWL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Find 5 trending MLOps repositories on GitHub this week and return name, description, stars and URL for each", "maxCredits": 100}'

Paste it into Import cURL and replace $FIRECRAWL_API_KEY with your key, or set up a Bearer credential. The response returns a job id right away instead of the results. Add a Wait node, then a second HTTP Request node that reads the job:

curl -s "https://api.firecrawl.dev/v2/agent/<jobId>" \
  -H "Authorization: Bearer $FIRECRAWL_API_KEY"

The job's status moves from processing to completed, and the results arrive in data. In our test run, data was still empty on the first completed poll and filled in a few seconds later, so loop back to the Wait node with an If node while status is processing or data is still empty. Send status: "failed" to an error branch instead of polling again. maxCredits caps the run at 100 credits; a run that hits the cap stops with status: "failed" and stopReason: "credit_limit_reached".

Expanding the workflow

With the research step configured, you can extend this workflow by adding nodes to:

  • Filter results for specific criteria
  • Send email or Slack notifications
  • Store data in databases or spreadsheets
  • Compare current trends with historical data

Method 2: self-hosted n8n with the Firecrawl node

Some automation workflows require complete data control and strict security compliance. Financial institutions processing sensitive customer data, healthcare organizations handling patient records, or government agencies managing classified information often cannot use cloud-based automation platforms. These scenarios call for self-hosted n8n.

When to use this method: Choose self-hosted n8n when you need complete control over where your workflows run, or when security requirements rule out cloud automation platforms. This method fits teams with existing infrastructure who are comfortable managing self-hosted applications.

Installing self-hosted n8n

If you have Docker Desktop installed, setting up n8n takes two commands.

First, create a volume to persist your workflow data:

docker volume create n8n_data

Then launch the n8n instance:

docker run -it --rm --name n8n -p 5678:5678 -v n8n_data:/home/node/.n8n docker.n8n.io/n8nio/n8n

Your n8n instance will be available at http://localhost:5678. The interface looks the same as the cloud version. For production deployments, see the n8n self-hosting documentation.

Registering your instance

Self-hosted n8n runs as the free Community edition. You can register it with your email for a free license key that unlocks a few extra features. Register during your first account setup, or later under Settings > Plan & usage > Unlock.

Registration unlocks:

  • Folders: organize workflows into folders
  • Debug in editor: copy and pin execution data while you work on a workflow
  • Custom execution data: save, find, and annotate execution metadata

Paid features such as SSO, environments, projects and external secrets need a Business or Enterprise license. The n8n editions page has the current breakdown.

Installing the Firecrawl node

Only the instance owner or an admin can install nodes, and they do it once for everyone on the instance:

  1. Open a workflow and click + to open the nodes panel
  2. Search for "Firecrawl" and select the node with the verified badge
  3. Click Install node

Firecrawl verified node details in a self-hosted n8n nodes panel, with the Install node button

The verified Firecrawl node in a self-hosted n8n 2.47.3 instance, before installation

If your instance installs community packages from npm instead, go to Settings > Community nodes, click Install, and enter the package name @mendable/n8n-nodes-firecrawl.

Connecting your API key

  1. Add the Firecrawl node to a workflow and click Connect to Firecrawl in the Credential field
  2. Paste your Firecrawl API key from the API keys page into API Key
  3. Check that Base URL is https://api.firecrawl.dev/v2, then click Save
  4. Set Resource to Scraping and Operation to Scrape URL, enter a URL, and click "Execute step"

Firecrawl account credential dialog in self-hosted n8n with Base URL and API Key fields

The credential is saved for every Firecrawl node in your workflows.

Now that you understand both setups, try building a simple workflow that scrapes a news website or documentation page with the Firecrawl node. The same steps work on n8n Cloud and on your self-hosted instance.

How do you fix common Firecrawl node errors in n8n?

Most Firecrawl node errors in n8n come from an old credential, a wait limit, or a job that returns an ID instead of results. These are the fixes the Firecrawl n8n docs give:

  • Agent, Browser or Interact returns a 404 ("The resource you are requesting could not be found"): the credential still points at https://api.firecrawl.dev/v1. Open it and set Base URL to https://api.firecrawl.dev/v2.
  • Extract Data With AI Agent returns timeout: the job ran past Max Wait Time (300 seconds by default, 600 at most). Read the result later with Get Agent Status and the returned jobId, or use Start AI Agent Job (Async).
  • Crawl Website returns only a job ID: crawls run as jobs. Add a second Firecrawl node with Get Crawl Status to read the scraped pages. Batch Scrape URLs works the same way with Get Batch Scrape Status.
  • The Scrape URL step has no Profile field: update to the latest Firecrawl node, or pass the profile object under Additional Fields > Custom Properties (JSON).

Using the Firecrawl /agent endpoint in n8n

Firecrawl /agent is a research agent that finds sources, chooses what to fetch, and returns structured results from a single prompt. You don't need to supply a specific URL or a schema up front. The agent decides which pages to visit and how to structure the output.

With the agent, you can run open-ended research tasks from a single prompt: it finds sources, navigates across pages, and returns structured results without requiring a URL list or schema. This makes it a good fit for enrichment workflows (for example, filling spreadsheet columns with founder names, roles, funding rounds, or product categories) and for tasks where the required sources aren't known ahead of time.

In n8n, the Firecrawl node gives you both variants. Extract Data With AI Agent waits for the result in one step, and Start AI Agent Job (Async) returns a job ID right away so you can fetch the result later with Get Agent Status. The agent pairs well with downstream nodes that filter, transform, or route the structured output to CRMs, sheets, or data warehouses.

Parallel Agents let you run hundreds or thousands of /agent prompts at once, so enrichment and research workflows that used to take hours can finish in minutes. This helps most in batch workflows like lead enrichment, market research, or catalog cleanup where you want structured results at scale.

The agent also works alongside other Firecrawl tools in n8n AI Agent workflows. Combine search, scrape, and agent steps to handle both predictable extraction and more exploratory research in one automation.

Using the Firecrawl /interact endpoint in n8n

Firecrawl /interact extends scraping into full browser control. After scraping a page, you stay in that session and tell Firecrawl what to do next: click a button, fill a form, move to the next page, or read data that only appears after an action.

This is useful in n8n workflows that hit paginated results, trigger search forms, or need to act on dynamic content before extracting it. In the Firecrawl node, set Resource to Interact and use:

  • Execute Interaction: runs a prompt, or Playwright or Bash code, against a scraped page, using the scrapeId from the Scrape URL output (data.metadata.scrapeId)
  • Stop Interaction: ends the session for that scrapeId
  • Get Interact Context: returns usage instructions for AI agents that call these operations as tools

Sessions expire after 10 minutes by default, or after 5 minutes without activity, so you can chain several interaction steps, including browser automation actions, within one run. Firecrawl bills Interact per browser minute: 2 Firecrawl credits a minute for code-only sessions and 7 when you use a prompt, with a one-minute minimum, so run Stop Interaction when the workflow is done (billing docs).

Practical example: chatbot with web access

Let's build a complete example that puts n8n and Firecrawl together: an AI chatbot that searches the web in real time with Firecrawl /search.

Read More: How Credal Extracts 6M+ URLs Monthly to Power Production AI Agents with Firecrawl

This example works on n8n Cloud and on self-hosted n8n.

Building the core workflow

To get started, create a workflow in n8n. The foundation requires two components that work together to create an interactive AI experience:

  1. "When chat message received" trigger: creates an interactive chat interface
  2. "AI Agent" node: connect this to your trigger node

The AI Agent node implements a ReAct (reason + act) pattern. This approach lets your chatbot think through problems step by step and use tools when needed to find answers, including Firecrawl search, scrape, and /agent calls.

Configuring the AI agent

Before the chatbot can work, the AI Agent needs three connections configured. Each connection handles a different part of the chatbot.

Chat Model Setup: Start by clicking the "+" next to "Chat Model" and adding your preferred language model provider. Configure it with your API credentials and select a model like gpt-4o-mini for cost-effective performance.

Memory Configuration: Next, add "Simple Memory" which stores conversation history on n8n's servers. This allows the chatbot to remember previous messages in your conversation. For production workflows, consider external databases like MongoDB, Redis, or PostgreSQL for better persistence and control.

Once configured, test your basic chatbot by clicking "Open chat" and sending a few messages. This testing step serves two purposes: it verifies the setup works and exposes the chatInput field that we'll need for the Firecrawl integration.

Adding web search

With the basic chatbot working, you can give it web search for AI agents through Firecrawl /search. Click the "Tools" edge and add the Firecrawl node with Resource set to Map & Search and Operation set to Search Web. The AI Agent can then decide when to call it.

When the node runs as a tool, you can let the model fill in the query. Click "Let the model define this parameter" next to the query field, which uses n8n's $fromAI() function. The node's README notes that tool use needs n8n 1.79.0 or later, and on self-hosted n8n it asks you to set N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true.

If you'd rather use the HTTP Request node as the tool, import this cURL command and map the query value to chatInput:

curl -s -X POST "https://api.firecrawl.dev/v2/search" \
  -H "Authorization: Bearer $FIRECRAWL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "n8n release notes", "limit": 3}'

Results come back under data.web, one object per result. This is a trimmed response from that request:

{
  "success": true,
  "data": {
    "web": [
      { "url": "https://docs.n8n.io/changelog/release-notes", "title": "Release notes | Changelog" },
      { "url": "https://github.com/n8n-io/n8n/releases", "title": "Releases · n8n-io/n8n" },
      { "url": "https://docs.n8n.io/changelog", "title": "Changelog" }
    ]
  },
  "creditsUsed": 2
}

Configure the Tool Description: The AI Agent needs instructions about when to use this tool. In the node description field, add:

Use this tool to search the web for real-time information.

This description guides the AI Agent's decision about when to search the web.

Testing your web-enabled chatbot

Return to the canvas and open the chat interface. Your chatbot can now pull live web information through Firecrawl /search.

The AI Agent evaluates each user question and decides when to search the web. This gives users current information that goes beyond the model's training data. You can adapt this same pattern for any workflow where AI agents need fresh web content or specific domain knowledge.

Can you connect the Firecrawl MCP server to n8n instead?

Yes. n8n's MCP Client Tool node lets an AI Agent call the tools of an external MCP server, with Bearer authentication. The Firecrawl MCP server takes API-key connections at https://mcp.firecrawl.dev/v2/mcp with your key as a Bearer token, and the agent then picks Firecrawl tools by itself. One n8n user shared a build of this setup on Reddit, using it to track competitors' Amazon store pages. The Firecrawl node gives you more control over each call, and the MCP route lets the model choose the tools.

Conclusion

This tutorial showed how to set up the Firecrawl node on n8n Cloud and on self-hosted n8n, when to use the HTTP Request node, and how to build a chatbot with live web data. On n8n Cloud, Connect to Firecrawl gets you running with 10,000 free Firecrawl credits. On self-hosted n8n, an admin installs the same verified node once for the whole instance.

For most workflows, the Firecrawl node is the place to start: it covers search, scrape, interact, crawl, map and agent, and it plugs into n8n's AI Agent node as a tool. Keep the HTTP Request node for API parameters the node doesn't expose, or when you want to paste a cURL command from the docs.

Continue your automation journey

Use Firecrawl in your agent

Add Firecrawl to ChatGPT, Claude, Cursor and more in one click.

Frequently Asked Questions

What is the Firecrawl node for n8n?

The Firecrawl node is a verified n8n node, published as @mendable/n8n-nodes-firecrawl and built and maintained by Firecrawl. It brings Firecrawl Search, Scrape, Crawl, Map, Interact and Agent into n8n workflows, and the same node runs on n8n Cloud and on self-hosted n8n.

How do I connect Firecrawl to n8n Cloud?

Add the Firecrawl node from the nodes panel and click Connect to Firecrawl. n8n creates a Firecrawl team linked to your n8n account and adds 10,000 free Firecrawl credits. You can also paste an API key from your Firecrawl dashboard into a Firecrawl credential instead.

What is included in the current n8n Cloud credit offer?

The launch offer that included a free Hobby plan and 100,000 credits has expired. Connecting through n8n Cloud now gives you 10,000 free Firecrawl credits without a Hobby plan upgrade. Firecrawl Scrape costs 1 credit per page, Firecrawl Search costs 2 credits per 10 results, and Firecrawl Interact bills 2 credits per browser minute for code-only sessions or 7 when you use a prompt.

Does the Firecrawl node support /agent?

Yes. The Agent resource has Extract Data With AI Agent, which waits for the result, and Start AI Agent Job (Async), which returns a job ID that you read later with Get Agent Status. You describe the data you need in a prompt and can add optional URLs and a JSON schema.

Why does the Firecrawl node return a 404 for Agent or Interact?

Credentials created with older versions of the node can still point at the v1 API. Operations that only exist in v2, such as Agent, Browser and Interact, return a 404 against v1. Open the Firecrawl credential in n8n and set Base URL to https://api.firecrawl.dev/v2.

What can I do with Firecrawl /interact in n8n?

Interact keeps working with a page after Scrape URL. You can click buttons, fill forms and page through results with a natural language prompt or Playwright code, using the scrapeId from the scrape output. Run Stop Interaction when the workflow is done so the browser session closes.

When should I use the HTTP Request node instead of the Firecrawl node?

Use the HTTP Request node when you need an API parameter the Firecrawl node doesn't show, or when you want to import a cURL command from the Firecrawl API reference as-is. For most workflows the Firecrawl node covers what you need, including Agent, Interact, batch scrape and crawl status.

Does the Firecrawl node work on self-hosted n8n?

Yes. An instance owner or admin searches for Firecrawl in the nodes panel and installs it once, and every member of the instance can then use it with a Firecrawl API key. Instances that install packages from npm can add @mendable/n8n-nodes-firecrawl under Settings, Community nodes.

Can I use the Firecrawl node as a tool in an n8n AI Agent?

Yes. Connect the Firecrawl node to the Tools input of an AI Agent node and the model decides when to search or scrape. The node's README lists n8n 1.79.0 or later for tool use, and on self-hosted n8n it asks you to set N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE to true.

How do Parallel Agents make /agent faster at scale?

Parallel Agents let you run hundreds or thousands of Firecrawl /agent prompts at once, which suits batch enrichment and research. The Parallel Agents announcement on the Firecrawl blog explains how it works.

//
Get started
//
Ready to build?
Start getting web context for free and scale seamlessly as your project expands. No credit card needed.
Are you an AI agent? See setup options