Firecrawl

• ▸Scrape - convert a single page into markdown, structured JSON, or a screenshot
• ▸Crawl - follow links across an entire site and return content for every page
• ▸Map - return the full list of indexed/discoverable URLs on a site without scraping content
• ▸Search - query the web and optionally return full page content for results, not just links
• ▸Extract - pull structured data from pages using natural-language prompts / a schema (the vendor now positions Agent as the successor to /extract)
• ▸Parse - upload a PDF, document, spreadsheet, or HTML file and get markdown or structured output
• ▸Interact - drive a live browser session (click, fill forms, navigate) via AI prompts to reach dynamic content
• ▸Monitor - track a page or site for changes over time
• ▸Agent (preview) - an autonomous research agent that browses and extracts structured data on its own

Authentication

API key (Bearer token), e.g. Authorization: Bearer fc-xxxxxxxx; no key required for basic/rate-limited use

Base URL

https://api.firecrawl.dev

Rate limits

Per-team requests per minute by plan: /scrape, /map, and /search allow 10 (Free), 100 (Hobby), 500 (Standard), 5,000 (Growth), 10,000 (Scale); /crawl and /agent allow 2, 20, 100, 1,000, 2,000. Concurrent browsers: 2 (Free), 5 (Hobby), 25 (Standard), 50 (Growth), 100+ (Scale/Enterprise).

Notable endpoints

• POST /v2/scrape - extract a single page as markdown, HTML, JSON, or screenshot
• POST /v2/crawl - asynchronous crawl of an entire site
• POST /v2/map - return the list of URLs on a site
• POST /v2/search - query the web and optionally scrape the results
• POST /v2/interact - create a live browser session
• POST /v2/agent - start an agentic data-extraction task
• POST /v2/parse - upload and parse a file
• POST /v2/extract - LLM-based structured extraction from pages (Agent is positioned as its successor)

MCP tools exposed

firecrawl_scrapefirecrawl_mapfirecrawl_searchfirecrawl_parsefirecrawl_crawlfirecrawl_check_crawl_statusfirecrawl_agentfirecrawl_agent_statusfirecrawl_interactfirecrawl_interact_stopfirecrawl_developer_searchfirecrawl_research_search_papersfirecrawl_research_inspect_paperfirecrawl_research_related_papersfirecrawl_research_read_paperfirecrawl_monitor_createfirecrawl_monitor_listfirecrawl_monitor_getfirecrawl_monitor_updatefirecrawl_monitor_runfirecrawl_monitor_deletefirecrawl_monitor_checksfirecrawl_search_feedbackfirecrawl_feedbackfirecrawl_monitor_checkfirecrawl_find_toolsfirecrawl_credit_usage

Machine-readable spec

OpenAPI spec retrieved

Published by the vendor themselves — not community forks.

• firecrawl-mcp-server7504 stars

  Official Firecrawl MCP Server - adds web scraping and search to Cursor, Claude, and other LLM/MCP clients.

• cli637 stars

  Official CLI and Agent Skill for Firecrawl - adds scrape, search, and browsing capabilities to AI agents.

• firecrawl-claude-plugin226 stars

  Firecrawl plugin for Claude Code - web scraping and crawling.

• skills107 stars

Checked 2026-09-23 against their pricing page. Vendors change pricing without notice.

Firecrawl is a developer-focused API product (by the team behind Mendable) that packages web scraping, crawling, and browser automation into a single credit-metered API and official MCP server, aimed at teams building AI agents, RAG pipelines, lead-enrichment tools, and research assistants that need live web data without maintaining their own scraping infrastructure. The vendor claims 1.25M+ developers and 150,000+ companies use it, naming teams at Apple, Canva, and Lovable, and says it has served more than 5 billion requests; the core `firecrawl/firecrawl` GitHub repo (AGPL-3.0) has 183K+ stars. On 2026-09-22 the vendor announced a $75M Series B alongside Alexandria, a source-discovery layer for AI agents.

Developers and AI teams building agents, RAG pipelines, lead enrichment, competitive intelligence, content generation, price monitoring, or chatbots that need reliable, LLM-ready web data via API or MCP - from solo builders on the free tier up to enterprises needing zero-data-retention and SSO.

Teams without engineering resources to call an API or wire up an MCP client should note that the official material checked on 2026-08-05 emphasized a data-extraction API and open-source app examples rather than a point-and-click GTM application. High-volume scraping teams should also model per-credit cost because credits are consumed by page, search, or browser minute and do not roll over on the retained self-serve plans.

Infrastructure for fetching web pages at scale — proxies, headless browsers, and crawl-to-markdown services that agents can call directly.

Category: Scraping and browser infrastructure. Membership is decided by 3 stated tests, not by category tags — see them and the full comparison.

ApifyAPIA marketplace of pre-built scrapers on top of the infrastructure, so common targets need no code at all.Bright DataBrowser UseAPIBrowserbaseAPIMCPOxylabsAPIScrapingBeeAPI

• Company headcount or valuation figures (funding is now confirmed: $75M Series B, changelog 2026-09-22)