Firecrawl MCP. Scrape and crawl websites into clean Markdown data.
Gemini Works with every AI agent you already use
…and any MCP-compatible client
Just plug in your AI agents and start using Vinkius.
Firecrawl. Scrapes and crawls entire websites into clean, structured markdown using a single API call. It handles JavaScript rendering, automatically excludes boilerplate content like headers and footers, and allows your AI agent to programmatically discover and ingest full knowledge bases from any root URL.
What your AI agents can do
Cancel active crawl
Stops a crawl job that is currently running using a provided job ID.
Get api usage
Checks and reports your current Firecrawl credit usage and remaining limits.
Get crawl status
Retrieves the current status (running, completed, failed) for a specific crawl job ID.
The agent sends a URL, and the server returns the content as clean, structured markdown, automatically cleaning up boilerplate elements.
The agent inputs a root domain, and the server lists every reachable link on that site without downloading the content.
The agent inputs a root URL and a depth limit, and the server initiates a background job that systematically discovers and scrapes all linked pages.
The agent sends a job ID, and the server returns the current status of the crawl, letting you know if it's running, finished, or failed.
The agent asks for usage details, and the server returns your remaining Firecrawl credits and usage history.
The agent provides an active job ID, and the server sends a signal to terminate the crawling process immediately.
Ask AI about this MCP
Supported MCP Clients
OAuth 2.0 Compatible Gemini Waiting for input…
Firecrawl MCP Server: 6 Tools for Web Data Extraction
These six tools let your AI agent manage the full lifecycle of web data: from mapping site structures to scraping individual URLs and monitoring job status.
Make your AI actually useful.
Add this MCP to Claude, Cursor, or Windsurf and your AI stops guessing. It gets real tools to look things up, take action, and handle the stuff you keep doing by hand.
Start using Firecrawl on Vinkius019dd0f0cancel active crawl
Stops a crawl job that is currently running using a provided job ID.
019dd0f0get api usage
Checks and reports your current Firecrawl credit usage and remaining limits.
019dd0f0get crawl status
Retrieves the current status (running, completed, failed) for a specific crawl job ID.
019dd0f0map website structure
Discovers and lists all accessible URLs on a domain without downloading the content of the pages.
019dd0f0scrape url
Converts the content of a single specified URL into clean, structured markdown format.
019dd0f0start crawl
Initiates a recursive crawling job on a root URL, returning a job ID for tracking.
Choose How to Get Started
Build a custom MCP for your own tools, or connect a ready-made integration from our catalog.
Build Your Own
Turn any API into an MCP. Import a spec, define Agent Skills, or deploy with MCPFusion.
- Import from OpenAPI, Swagger, or YAML specs
- Create Agent Skills with progressive disclosure
- Deploy to edge with MCPFusion framework
- Built in DLP, auth, and compliance on every call
- Real time usage dashboard and cost metering
- Publish to catalog or keep private
Make Your AI Do More
Start with Firecrawl, then connect any of our 4,800+ other servers whenever your AI needs more. One click, no limits.
- Use this MCP plus 4,800+ others, all in one place
- Add new capabilities to your AI anytime you want
- Every connection is secured and compliant automatically
- Track usage and costs across all your servers
- Works with Claude, ChatGPT, Cursor, and more
- New servers added to the catalog every week
Independent Platform Disclaimer: Vinkius is an independent platform and is not affiliated with, endorsed by, sponsored by, verified by, or otherwise authorized by Firecrawl. All third-party trademarks, logos, and brand names are the property of their respective owners. Their use on this website is strictly for informational purposes to identify service compatibility and interoperability.
VINKIUS INFRASTRUCTURE
Cloud Hosted
Managed infra
V8 Isolated
Sandboxed per request
Zero-Trust Proxy
No stored credentials
DLP Enforced
Policy on every call
GDPR Compliant
EU data residency
Token Compression
~60% cost reduction
Works with Claude, ChatGPT, Cursor, and more
The Model Context Protocol standardizes how applications expose capabilities to LLMs. Instead of operating in isolation, your AI gains direct access to external platforms, live data, and real-world actions through secure, standardized connections.
This server provides 6 capabilities that interface natively with Claude, ChatGPT, Cursor, and any MCP client. No middleware. No custom integration required.
Manually gathering web data is a nightmare of copy-pasting.
Today, you open a website, copy the text, paste it into Notion, and then manually cut and paste the relevant sections into your research document. You spend minutes filtering out the boilerplate—the headers, the footers, the related articles widgets—just to get to the actual content. It's slow, tedious, and you always miss something.
With Firecrawl, you just point your agent at the URL. It handles the whole process, automatically stripping out the junk content and giving you the core, clean text you need. It's done in a single conversation turn.
Firecrawl MCP Server: Extract clean content and structure.
You no longer have to switch between a browser, a text editor, and a spreadsheet. The agent handles the entire data pipeline: first, it uses `map_website_structure` to see the site map, then `start_crawl` to get the content, and finally, it presents the structured data directly back to your chat window.
The difference is the workflow. You don't manage the crawl, the state, or the API keys. Your AI client manages it all. You just ask the question.
What you can do with this MCP connector
Firecrawl lets your AI agent scrape and crawl entire websites, spitting out clean, structured markdown. You'll use this server to ingest full knowledge bases from any root URL, and it handles JavaScript rendering and stripping out all the boilerplate junk like headers and footers.
To grab a single URL, your agent sends the address and gets the content back as structured markdown, automatically cleaning up the fluff. You can also map a site's internal structure by giving it a root domain; the server lists every reachable link without actually downloading the content. If you wanna crawl a whole site, your agent inputs a root URL and a depth limit, and the server starts a background job that systematically discovers and scrapes all linked pages.
When you need to track a crawl job, your agent sends a job ID, and the server returns the current status: running, finished, or failed. You can also check your API usage by asking for details, and the server spits out your remaining Firecrawl credits and usage history. If a crawl job gets out of hand, your agent can provide an active job ID, and the server sends a signal to terminate the process immediately.
019dd0f1-1cfd-731d-bbff-c84b7b8f732b 
How Firecrawl MCP Works
- 1 Subscribe to the Firecrawl server and retrieve your API Key from the Firecrawl dashboard.
- 2 Give your AI agent the target URL and the desired action (e.g., 'Scrape this page' or 'Crawl this site').
- 3 The agent executes the tool call, and the server returns the structured data (Markdown, job ID, or usage metrics) to your agent for processing.
The bottom line is: you tell your AI agent what web data you need, and it uses Firecrawl to execute the scrape or crawl, giving you clean, structured output.
Who Is Firecrawl MCP For?
This is for the data-intensive roles: the AI developer building RAG pipelines, the research team needing market data, and the content creator who needs to convert articles fast. If your job involves gathering information from multiple web sources, this is what you need.
Builds Retrieval Augmented Generation (RAG) applications by programmatically feeding clean, structured web data into the system.
Automates gathering competitive intelligence and market analysis from multiple, disparate websites without manual copy-pasting.
Converts full-article web pages into clean Markdown and captures visual screenshots to maintain a content archive.
What Changes When You Connect
- Structured Data Output: Forget messy HTML.
scrape_urlconverts complex web pages into clean, structured Markdown, making the data instantly usable by your LLM. - Deep Site Discovery: Need more than just one page?
start_crawlhandles recursive crawling, systematically finding and scraping every linked subpage to build a full knowledge base. - Planning Before Execution: Don't guess what's on the site. Use
map_website_structurefirst to get a full list of reachable URLs, letting you plan your data gathering before spending credits. - Full Process Visibility: Use
get_crawl_statusandget_api_usageto monitor jobs in real-time. You always know if the crawl is running, finished, or if you're running low on credits. - Visual and Textual Record: Capture a full-page screenshot alongside the structured text using the agent's visual capture capability. You get both the raw visual context and the clean text data.
- Stop and Control: If a crawl goes wrong or you change your mind,
cancel_active_crawllets you shut down the job immediately.
Real-World Use Cases
Building a RAG Knowledge Base
A developer needs to index all documentation from a vendor site. Instead of manually scraping 50 pages, they ask their agent to use map_website_structure first, then start_crawl on the root URL. The resulting job ID feeds into their RAG pipeline, providing a complete knowledge base.
Competitive Analysis
A market researcher needs to track competitor pricing across 10 product pages. They use scrape_url on each specific URL, ensuring that every piece of extracted content is clean Markdown, allowing for easy comparison and structured data analysis.
Archiving an Article
A content creator finds a great article and needs to save it for later. They ask their agent to use scrape_url, which extracts the main text and automatically strips out the distracting sidebars and ads. They also capture a full-page screenshot for context.
Auditing a Website's Links
A site auditor needs to know every possible internal link on a large corporate site. They use map_website_structure to get a complete list of all reachable URLs, without wasting time or credits on content extraction.
The Tradeoffs
Scraping everything at once
Calling start_crawl on a massive, unknown site without checking the scope or depth limit. This risks hitting rate limits, wasting credits, and generating a massive, unmanageable data dump.
→
First, use map_website_structure to scope the site. Then, use start_crawl with a defined depth. If you only need content from one specific page, bypass the crawl and use scrape_url instead.
Relying on raw HTML output
Taking the raw output from a simple scraper and feeding it to the LLM. The LLM has to filter out headers, footers, and navigation menus, which introduces noise and weakens the context.
→
Always use scrape_url. It guarantees the output is clean, structured Markdown, focusing only on the core content needed for the LLM.
Ignoring job status
Initiating a crawl with start_crawl and then forgetting to check the progress. You don't know if the job is still running, stuck, or if it failed due to an internal error.
→
After calling start_crawl, immediately use get_crawl_status with the returned job ID. This confirms the job started and lets you track it until completion.
When It Fits, When It Doesn't
Use this if your task is data acquisition from the web. Specifically, if you need to build a knowledge base, extract core article text, or map out a site's structure. You absolutely must use this if your process relies on clean, structured markdown output.
Don't use this if you just need to process data that is already structured (like a CSV or database query). In those cases, a simple data connector is better. Also, don't use this if you only need a single, static image; use a dedicated image capture tool instead. Firecrawl is for getting text and structure from websites, period.
Common Questions About Firecrawl MCP
How do I use Firecrawl with my AI client to scrape a single page? +
You use the scrape_url tool. Just give your agent the URL you want. It handles the JavaScript rendering and outputs clean Markdown, making the content ready for analysis right away.
Can Firecrawl crawl an entire website recursively? +
Yes. You use the start_crawl tool. It initiates a background job and returns a job ID. You then use get_crawl_status to monitor the progress until the entire site is indexed.
What is the difference between `map_website_structure` and `start_crawl`? +
map_website_structure only discovers links; it doesn't download content. start_crawl executes the crawl and downloads the actual page content. Use mapping to plan, and crawling to execute.
How do I check if my Firecrawl API usage is over my limit? +
Run the get_api_usage tool. This tells you your remaining credits and helps you manage your budget before you run out of data extraction capacity.
Can Firecrawl capture screenshots while crawling? +
Yes, the agent can capture full-page screenshots of any URL. This adds a visual record to your data set, giving you context beyond just the text.
How do I manage an ongoing crawl using the `cancel_active_crawl` tool? +
You call cancel_active_crawl with the job ID. This immediately stops the crawl job, preventing further processing and saving credits. You can't restart a canceled job; you'll need to initiate a new crawl.
What is the difference between `scrape_url` and `map_website_structure`? +
scrape_url converts a single URL into clean, Markdown-ready content. map_website_structure simply discovers all reachable links on a site without extracting any content. Use map_website_structure first if you only need a site map.
How do I monitor my job progress after running `start_crawl`? +
You use get_crawl_status with the job ID returned by start_crawl. This tells you if the crawl is running, paused, or complete. It's the only way to track the real-time status of a background job.
How do I find my Firecrawl API Key? +
Log in to your Firecrawl dashboard, and navigate to the API Keys section to copy your unique token.
Can I scrape content excluding headers and footers? +
Yes! The scrape_url tool includes an onlyMainContent parameter. When set to true, Firecrawl uses AI to extract only the core article or page content.
How long does a recursive crawl take? +
Crawl time depends on the site size and depth. Use the get_crawl_status tool to monitor progress and retrieve results once the job is complete.
Use it with your favorite AI tools
Connect this server to Cursor, Claude, VS Code, and more.