Firecrawl
by Firecrawl
Turn any website into clean, LLM-ready data with one API call
Firecrawl is an open-source API that scrapes, crawls and searches the web and returns clean markdown, HTML or structured JSON ready to feed an LLM. It handles the parts that break naive scrapers — JavaScript rendering, rotating proxies, rate limits and bot walls — so engineering teams building RAG pipelines and agents stop maintaining scraper infrastructure.
Firecrawl is a web data API that converts arbitrary websites into clean, LLM-ready output. Its three core endpoints are Scrape, which turns a URL into markdown, HTML, a screenshot or structured JSON; Crawl, which walks a whole site; and Search, which queries the web and returns full page content rather than snippets, with Map for URL discovery, Batch Scrape for bulk jobs, Extract for schema-shaped output and an autonomous Agent mode alongside them. The value is in what it absorbs: JavaScript-heavy pages, rotating proxies, rate limits and bot walls are handled by the service, and the company claims coverage of 96% of web pages at a P95 latency of 3.4 seconds. The project is open source and self-hostable under AGPL-3.0, with SDKs under MIT for Python, Node.js, Go, Java, Elixir, Rust, Ruby, .NET and PHP, plus a CLI, MCP servers and connectors for Zapier, n8n and Lovable. Firecrawl was co-founded in 2022 by Caleb Peffer (CEO), Nicolas Silberstein Camara (CTO) and Eric Ciarla (CMO), and grew out of their earlier product Mendable, an AI chat-over-docs tool used by teams at Snapchat, MongoDB and DoorDash; the team concluded the real opportunity was upstream, in getting web data to models at all. It announced a $14.5 million Series A on 19 August 2025 led by Nexus Venture Partners with participation from Y Combinator and Shopify CEO Tobias Lütke, at which point it reported 350,000 signed-up developers, nearly 50,000 GitHub stars, 15x growth over the prior year, profitability, and customers including Shopify, Replit, Zapier and several large hedge funds. Pricing is published and starts free at 1,000 credits per month.
An engineering team building a RAG pipeline or a web-browsing agent that is currently maintaining its own Playwright and proxy stack and losing time to broken selectors and bot walls.
Clean markdown or structured JSON from any URL through one API, with JavaScript rendering, proxies and rate limits handled, so no scraper infrastructure has to be owned.
At a Glance
- Category
- Developer Tools
- Pricing
- Freemium, Subscription, Usage-based
- Target Market
- CTOs, Enterprise Developers, Data Engineers, AI Engineers, Startup Founders
- Deployment
- Cloud-first, Open-source, Self-hosted, API-based
- Founded
- 2022
- Customers
- 350,000 signed-up developers as of August 2025; customers include Shopify, Replit and Zapier
Key Features
- ✓Scrape endpoint
Converts a single URL into markdown, HTML, a screenshot or structured JSON, already cleaned of navigation and boilerplate for LLM input.
- ✓Crawl and Map
Walks an entire site discovering and fetching every reachable page, or returns just the URL map for selective fetching.
- ✓Search with full content
Queries the web and returns whole page bodies rather than snippets, which is what an agent actually needs to reason over.
- ✓Structured extraction
Takes a schema and returns matching JSON from unstructured pages, removing the per-site parsing code teams otherwise write.
- ✓Anti-bot and JS handling
Rotating proxies, rate-limit management and JavaScript rendering are handled by the service, claimed to cover 96% of web pages.
- ✓Open source and self-hostable
The core is AGPL-3.0 with MIT-licensed SDKs, so the service can be run in-house where data cannot leave the network.
- ✓Nine SDK languages plus MCP
Python, Node.js, Go, Java, Elixir, Rust, Ruby, .NET and PHP SDKs, plus a CLI and MCP servers for agent frameworks.
Capabilities
Use Cases
- •Building a documentation RAG index
Crawl a vendor's entire docs site into clean markdown and embed it without writing any HTML parsing code.
- •Web-browsing agents
Give an agent a search-and-read capability that returns full page content, so it can reason over sources rather than snippets.
- •Competitive price and catalogue monitoring
Recrawl competitor product pages on a schedule and extract structured pricing fields into a database automatically.
- •Financial and regulatory research
Pull filings, announcements and regulator publications into a pipeline where hedge funds already use it at scale.
- •Lead and firmographic enrichment
Extract company details from public sites into a defined JSON schema for a CRM enrichment workflow.
Ideal For
Best For
- ✓Building RAG knowledge bases from public documentation, product catalogues or regulatory sites
- ✓Giving AI agents a reliable web-read capability without writing per-site scrapers
- ✓Competitive and market monitoring that needs recurring crawls of sites with JavaScript-heavy rendering
- ✓Extracting schema-shaped structured data from unstructured pages using a defined JSON output
- ✓Replacing an in-house Playwright plus proxy stack that keeps breaking on layout and anti-bot changes
Not Ideal For
- ✗Very high-volume scraping on a tight budget — heavy users on Hacker News report shopping for cheaper alternatives, and one described falling back to Kagi's extract API plus Cloudflare's browser API as more cost efficient
- ✗Sites behind authentication or hard paywalls, which no scraping API can lawfully or reliably solve for you
- ✗Teams whose licensing policy excludes AGPL-3.0 for self-hosted deployments — the SDKs are MIT but the core service is not
- ✗Anyone needing the full hosted feature set on-premise, since self-hosting the open-source version does not include all cloud capabilities
Integrations
Deployment
Market & Ratings
350,000 signed-up developers as of August 2025; customers include Shopify, Replit and Zapier
Market Analysis
Pros
- ✓Removes the single most brittle part of a RAG or agent pipeline — JavaScript rendering, proxies and bot walls are the service's problem, not yours
- ✓Open source and self-hostable, so there is a credible exit if the hosted pricing stops working
- ✓Nine SDK languages plus CLI and MCP coverage make integration close to trivial in most stacks
- ✓Published, predictable pricing with a genuinely usable free tier at 1,000 credits a month
- ✓Company reports profitability and 350,000 developer signups, which lowers the vendor-survival risk that dogs this category
Cons
- ✗Cost at volume is the recurring complaint — one self-described heavy user on Hacker News asked outright about getting a better price, and another reported that Kagi's extract API with a Cloudflare browser API fallback 'looks to be more cost efficient than firecrawl'
- ✗The category is unusually crowded and commoditising fast; practitioners list Browser Use, Browserbase, Parallel and Exa as substitutable, and differentiation is a live question in those threads
- ✗Credit-based metering makes true cost hard to predict, since consumption varies by endpoint and page complexity rather than tracking a simple page count
- ✗The core service is AGPL-3.0, which some corporate legal policies will not accept for self-hosted deployment even though the SDKs are MIT
- ✗Self-hosting does not reproduce the full hosted feature set, so the open-source escape hatch is narrower than it first appears
Pricing
Free
$0
- ✓1,000 credits per month
- ✓2 concurrent requests
- ✓No credit card required
Hobby
From $16/mo
- ✓Billed yearly
- ✓5 concurrent requests
- ✓Higher monthly credit allowance
Standard
From $83/mo
- ✓Billed yearly
- ✓25 concurrent requests
- ✓Production credit allowance
Growth
From $333/mo
- ✓Billed yearly
- ✓50 concurrent requests
Scale
From $599/mo
- ✓Billed yearly
- ✓Up to 1,000,000 credits per month
- ✓100 concurrent requests
- ✓Extra credits at $397 per 350k
Enterprise
Contact for pricing
- ✓Custom credits and unlimited pages
- ✓Custom concurrency
- ✓Dedicated support and SLA
- ✓Zero data retention
- ✓SSO and advanced security
- ✓Bulk discounts
Pricing is fully published and metered in credits, with concurrency the other axis that separates tiers. The free tier gives 1,000 credits a month with 2 concurrent requests and no card; paid plans run $16, $83, $333 and $599 per month billed yearly, scaling to 1,000,000 credits and 100 concurrent requests, with overflow at $397 per extra 350,000 credits. Zero data retention, SSO, unlimited pages and an SLA are gated behind Enterprise. Credit consumption varies by endpoint, so model your own mix before comparing to a competitor's headline rate.
Security & Compliance
Connect
Sources
This page was written from 5 sources, 4 on domains other than firecrawl.dev.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Slack Code
Agentic coding in shared Slack channels, where the whole team can see the diff
Honeycomb
OpenTelemetry-native observability that traces AI agents and the systems they touch in one view
Tricentis Agentic Quality Engineering Platform
A governed fleet of testing agents that validates AI-written enterprise software before it ships
AlphaEvolve
Gemini-powered evolutionary agent that discovers faster algorithms for your hardest optimization problems