Firecrawl

Firecrawl is a web scraping API that turns websites into clean, structured data. The Firecrawl MCP server lets you search, scrape, crawl, and extract data from websites using natural language.

What Can It Do?

Search the web with optional scraping and source filtering
Scrape single URLs for content in markdown, HTML, or other formats
Map websites to get all URLs ordered by relevance
Crawl entire sites and extract content from multiple pages
Deep extract data by autonomously navigating and exploring links
Interact with pages using natural language prompts or browser code
Manage interactive browser sessions for persistent browser automation

Where to Use It

Add Firecrawl as a tool to any agent. The agent can then scrape and extract web data conversationally, choosing the right actions based on context.

To add an MCP tool to your agent:

Open your agent’s configuration
Click Add tools → Connect an app with MCP
Search for the integration and select it
Authenticate with your account

You can control which tools your agent has access to. After adding an integration, click on it to enable or disable specific tools based on what your agent needs.

Available Tools

Tool	Description	Credits
Search	Search the web and optionally scrape full page content. Returns results organized by source type (web, images, news).	8 per item
Scrape	Scrape a single URL and extract content in various formats. Supports persistent profiles to reuse browser state across sessions.	8
Map	Get all URLs from a website. Returns a list of URLs ordered by relevance.	1
Crawl	Crawl a website and extract content from multiple pages.	40
Get Crawl Status	Get the status and results of a crawl job.	8 per item
Batch Scrape	Scrape multiple URLs at once.	40
Get Batch Scrape Status	Get the status and results of a batch scrape job.	8 per item
Deep Extract	Autonomously navigate and extract data from websites based on a prompt. Unlike regular extract, this explores links and pages to find relevant data.	120
Get Deep Extract Status	Get the status and results of a deep extract job.	3
Interact	Interact with a previously scraped page using a natural language prompt or browser code.	14
Stop Interact	Stop an interactive browser session to release resources.	3
Create Interact Session	Start a standalone interactive browser session. Supports persistent profiles scoped per user.	4
Execute Interact Session	Run Playwright or agent-browser code in a standalone interact session.	4
Delete Interact Session	Stop a standalone interact session to release browser resources.	3
List Interact Sessions	List standalone interact sessions, optionally filtered by status.	3

The Gumloop-managed Firecrawl key supports searching, scraping, mapping, crawling, and extraction tools. Interactive browser tools — Interact, Stop Interact, Create Interact Session, Execute Interact Session, Delete Interact Session, and List Interact Sessions — require your own Firecrawl API key. Connect it in your Connectors page.

Example Prompts

Use these with your agent: Scrape a page:

Scrape this URL and get the main content as markdown

Search the web:

Search for "AI startup funding" and get the top 10 results

Map a website:

Get all URLs from example.com

Crawl a site:

Crawl example.com/blog with depth 2 and get all article content

Deep extract:

Extract pricing information from this SaaS website, exploring all relevant pages

Batch scrape:

Scrape these 5 URLs and get the main content from each

Interactive browser session:

Create a browser session, navigate to example.com, and click the login button

Troubleshooting

Issue	Solution
Agent not finding the right data	Ensure the URL is publicly accessible
Interactive tool requires API key	Interactive browser tools require your own Firecrawl API key. Connect it in your Connectors page
Action not completing	Check that you’ve authenticated and have sufficient Firecrawl credits
Unexpected results	The agent may chain multiple tools (e.g., mapping first, then scraping). Review the agent’s reasoning to understand its approach.
Tool not available	Verify the tool is enabled in your agent’s MCP configuration

Agents are smart enough to chain multiple API calls together. For example, asking “Get all blog posts from this site” will map the URLs first, then scrape each one. If results seem off, check the agent’s step-by-step reasoning.

Need Help?

Agents documentation for setup and best practices
Need help? Reach out to us
Contact support@gumloop.com for assistance

Use this integration directly in Claude or Cursor. Connect remotely via the Firecrawl MCP server using credentials from your Connectors page.

Get Started

Core Concepts

Enterprise Features

Connectors

What Can It Do?

Where to Use It

Available Tools

Example Prompts

Troubleshooting

Need Help?

​What Can It Do?

​Where to Use It

​Available Tools

​Example Prompts

​Troubleshooting

​Need Help?

What Can It Do?

Where to Use It

Available Tools

Example Prompts

Troubleshooting

Need Help?