Firecrawl is a web data API for AI applications and agent integration, converting search results, web pages, and documents into content usable by models. Suitable for teams developing research agents, knowledge bases, customer service bots, or data collection pipelines, it retrieves Markdown, structured JSON, screenshots, and metadata via queries or URLs, reducing the work of self-handling page scraping and format cleaning.
Firecrawl is a web data API for AI applications and agent integration, converting search results, web pages, and documents into content usable by models. Suitable for teams developing research agents, knowledge bases, customer service bots, or data collection pipelines, it retrieves Markdown, structured JSON, screenshots, and metadata via queries or URLs, reducing the work of self-handling page scraping and format cleaning.
It integrates search, full-site crawling, dynamic page operation, document parsing, and source monitoring into a single service, and can be used via MCP, command line, or developer kits in various languages. When data needs continuous updating, monitoring schedules can be set with notifications for meaningful changes; when data outside the website is needed, Alexandria can be used to connect specialized indexes and data providers. The project also provides open-source code, and the hosted version features proprietary infrastructure.
Features
Web search: Enter queries to get relevant URLs and full-page Markdown, integrating search and content reading into a single call.
Web scraping and structured output: Convert single pages into Markdown, HTML, screenshots, or metadata, and extract fields according to a specified JSON schema.
Full-site crawling: Follow links starting from a seed URL to collect an entire website or specified sections, with configurable depth, page limits, and path filtering.
Dynamic page handling: Automatically execute JavaScript, combined with smart waiting and proxy management, to read dynamically loaded website content.
Browser interaction: Click, type, scroll, and navigate pages using AI prompts or code, handling information that requires sequential operations to acquire.
Document parsing: Convert documents like PDF, DOCX, XLSX, and HTML into Markdown while preserving layout structure, using OCR on pages where needed.
Alexandria data integration: Access research indexes, government indexes, and licensed data providers to obtain structured records with sources and citable paper paragraphs.
Source Change Monitoring: Compares source snapshots based on a schedule, uses AI to judge whether changes match specified targets, and then notifies via webhook or email.
Multi-step Research Agent: In preview, the Firecrawl Agent receives natural language tasks, independently plans and calls search, extraction, scraping, and parsing tools, and returns structured results with citations.
Development and Agent Integration: Provides an official MCP server, command-line tools, and development kits in multiple languages, allowing applications and compatible agents to directly use web data capabilities.
Data Protection and Access Restrictions: Supports personal data masking; Enterprise plans additionally provide zero data retention, API key permissions, and IP restrictions.
Use cases
Research Product Development: Allows research agents to search public sources, parse reports and papers, and retain URLs and citation information for subsequent organization of research summaries.
Knowledge Base Updates: Scrapes technical documentation to build RAG data sources, then triggers content updates with change monitoring notifications.
E-commerce Data Collection: Extracts prices, inventory, and product fields from product pages for existing analysis workflows to compare different sources.
Competitor Tracking: Regularly monitors competitor features, prices, and announcement pages, and receives notifications when changes matching tracking targets appear.
Categories: Development and Prompt Engineering
Pricing
Free plan 1,000 credits per month; Hobby billed at US$19/month, or equivalent to US$16/month billed annually; Enterprise plan custom-quoted