Skip to content
MCP Server Directory logo
search-knowledge-discovery

Mcp Firecrawl Tool

Reviewed overview

MCP サーバー概要

What is the project about? MCP Firecrawl Tool is a small Node.js server that wraps Firecrawl APIs to scrape websites and extract structured data. It exposes two tools — scrape-website and extract-data — and optionally reports errors to Sentry.

How to use the project?

  • Install dependencies: npm install
  • Configure credentials in a .env or environment variables: set FIRECRAWL_API_TOKEN (required) and SENTRY_DSN (optional)
  • Start: npm start
  • Invoke tools with the provided CLI integration (mcp-inspector). Example usages:
    • scrape-website: provide url and optional output formats (markdown, html, text)
    • extract-data: provide urls, a prompt describing the data to extract, and a schema describing fields and types

Key features

  • Website scraping with multiple output formats (markdown, html, text)
  • Structured data extraction based on prompts and user-defined schemas
  • Schema support for strings, booleans, numbers, arrays and nested objects
  • Sentry integration for error tracking and performance monitoring
  • Simple CLI integration via mcp-inspector for easy automation

Use cases

  1. Extracting company or product information across multiple pages for research or enrichment
  2. Price monitoring and feature comparison scraping
  3. Building datasets from public web pages (lead gen, market research)
  4. Archiving or converting site content to markdown/html/text for downstream processing
  5. QA or automation tasks that need structured outputs from web pages

FAQ

  • Do I need an API token?

Yes. FIRECRAWL_API_TOKEN is required to call Firecrawl APIs.

  • Is Sentry required?

No. SENTRY_DSN is optional; enable it to collect error and performance logs.

  • What formats are supported by scrape-website?

Supported formats are markdown (default), html, and text.

  • How do I define what data to extract?

Use extract-data with a natural-language prompt describing the desired fields and provide a schema object where keys are field names and values are types (“string”, “boolean”, “number”, arrays like [“string”], or nested objects).

  • What happens on failures?

The server returns error messages for failed scrapes/extractions and will log errors to Sentry if configured.

  • Can I extract from multiple pages at once?

Yes. extract-data accepts an array of URLs to process in a single request.