Firecrawl
apiby firecrawl
Web scraping API built for AI. Scrape, crawl, search, and extract structured data from any URL. Handles JavaScript, anti-bot, and proxy rotation so you do not have to.
Key features
Free tier available, Standard at $83/mo
Teams that need to pull information from websites reliably and at scale. Competitor pricing, lead data, product catalogs, content monitoring. Firecrawl handles the technical complexity so you get clean, usable data back.
96% success rate across the web. Most pages come back with clean, usable content on the first try. The cloud version handles JavaScript-heavy sites, dynamically loaded content, and bot protection automatically.
What it does
Scrape: Single Page Extraction
Extract content from any URL in markdown, HTML, structured JSON, screenshots, links, images, branding data, or summary format. Handles JavaScript rendering and dynamic content automatically with sub second response times.
Learn moreCrawl: Full Site Recursion
Recursively follow links from a starting URL and scrape all subpages. Async job with polling, WebSocket streaming, and webhook support. Respects sitemaps with a default limit of 10,000 pages per crawl.
Learn moreMap: URL Discovery
Instantly discover all URLs on a website using sitemap scanning and link following, without scraping content. Supports search filtering to find specific pages before deciding what to scrape.
Learn moreSearch: Web Search with Scraping
Search the web and optionally scrape the full content of search results in a single call. Supports web, image, and news sources with location targeting and search operators.
Learn moreExtract v2: LLM Powered Structured Data with Pagination
Extract structured data from web pages using Pydantic schemas or natural language prompts. Returns typed JSON matching your schema, powered by LLM processing on the extracted content. Extract v2 adds pagination support, enabling extraction across multi page result sets automatically.
Learn moreFIRE-1 Agent: Autonomous Web Crawling
FIRE-1 is Firecrawl's autonomous web crawling agent, introduced in v1.11.0. Describe what data you need and the agent searches, navigates, and extracts it autonomously without requiring URLs upfront. Uses Spark 1 mini or Spark 1 pro models for complex multi source research tasks.
Learn moreMCP Server for AI Coding Tools
First class Model Context Protocol server that integrates directly with Claude Code, Cursor, Windsurf, VS Code, and other AI tools. One command setup gives AI agents full web scraping capability.
Learn moreBrowser Sandbox
Fully managed secure browser environments for AI agents. Create persistent CDP sessions, execute Python, JavaScript, or bash code, take screenshots, and interact with pages programmatically.
Pricing
Free
1,000 credits per month to test the API. No credit card required.
- 1,000 credits per month
- 2 concurrent requests
- All core endpoints (scrape, crawl, map, search)
- Playground access
- Community support
Hobby
For individual developers and small projects. 3,000 credits per month with basic support.
- 3,000 credits per month
- 5 concurrent requests
- All core endpoints
- Extra credits at $9 per 1,000
- Basic support
Standard
For growing teams and production workloads. 100,000 credits per month with 50 concurrent requests.
- 100,000 credits per month
- 50 concurrent requests
- All endpoints including Extract and Agent
- Extra credits at $47 per 35,000
- Standard support
Growth
For high volume production use. 500,000 credits per month with priority support.
- 500,000 credits per month
- 100 concurrent requests
- All endpoints including Agent and Browser Sandbox
- Extra credits at $177 per 175,000
- Priority support
Pros & Cons
Pros
- 96% success rate across the web. Most pages come back with clean, usable content on the first try. The cloud version handles JavaScript-heavy sites, dynamically loaded content, and bot protection automatically.
- Zero infrastructure to maintain. No proxy pools, no headless browsers, no CSS selectors to keep up to date. Pass a URL, get markdown back. That simplicity is the point.
- Six distinct scraping modes cover every scenario: single page, full site crawl, URL discovery, web search, schema-based extraction, and autonomous agent. Most scraping tools give you one or two of these.
- Output is optimized for LLMs. Clean markdown that chunks well for RAG, structured JSON that matches your schema, screenshots when you need visual context. No HTML cleanup step.
- Open source core with 127,000+ GitHub stars. You can self-host if you need data residency, though the cloud version has significantly more capability.
- FIRE-1 autonomous agent can research across the web without you specifying URLs. Describe what you need, it finds and extracts it. Genuinely useful for multi-source research tasks.
Cons
- Credits burn fast when you use add-ons. JSON extraction costs 5 credits per page instead of 1, enhanced proxy adds 4 more. A crawl that looks cheap at 1 credit/page can become expensive quickly.
- No pay-per-use option. Monthly subscriptions only, and unused credits do not roll over. If your usage is unpredictable, you are either overpaying for headroom or running out mid-month.
- The self-hosted version is missing the features that make the cloud version reliable: Fire Engine, stealth proxies, anti-bot bypass, actions, browser sandbox, and the FIRE-1 agent. It is more of a proof of concept than a production deployment.
- No social media scraping. Twitter, LinkedIn, Instagram, Facebook are all explicitly unsupported. If you need social data, you need a different tool.
- Authenticated pages and login walls require manual action scripting. No built-in CAPTCHA solving or 2FA handling. For anything behind a login, expect extra work.
- FIRE-1 Agent pricing is opaque. You get 5 free runs per day, then 'dynamic pricing' with no clear per-run cost. Hard to budget for.
How to get started
Sign up and get your API key
Create a free account at firecrawl.dev. No credit card required. You receive 500 free credits to test all endpoints. Get your API key from the dashboard.
Test in the playground
Before writing any code, try the interactive playground at firecrawl.dev/playground. Enter any URL and see the scrape, crawl, map, and extract results instantly. This is the fastest way to understand what Firecrawl returns.
Install an SDK and make your first scrape
Install the Python SDK (pip install firecrawl-py) or Node SDK (npm install @mendable/firecrawl-js). Initialize with your API key and call scrape() on any URL. The response includes clean markdown ready for your LLM.
Add to your AI agent via MCP or CLI
For AI coding tools like Claude Code or Cursor, run 'npx -y firecrawl-cli@latest init --all --browser' to install the Firecrawl skill and MCP server in one command. Restart your agent and it will have full web scraping capability.
Deep dive
Detailed guides with comparisons, tips, and visuals for each feature.
Scrape vs. Crawl vs. Map vs. Search vs. Extract vs. Agent
When to use each Firecrawl endpoint, how they work, what they cost, and how they fit together in a typical workflow.
How Firecrawl Connects to AI Tools
How Firecrawl works with AI assistants, coding tools, and chatbots to give them access to web data.
Open Source vs. Cloud: What You Get with Each
The differences between self hosting Firecrawl and using the cloud API, including which features are cloud only.
Links
Official
Documentation
Blog
Pricing
Get notified about updates
We'll email you when this tool's pricing or features change.
Last updated: 2026-06-01