How is CRW faster than Firecrawl?

CRW is built in Rust — not Node.js. That means no V8 overhead, no garbage collection pauses, and native async I/O. Average scrape latency is 142ms vs 300ms+ on Firecrawl.

Our Standard plan gives you 100,000 credits for $49/mo. Firecrawl charges $99/mo for the same volume. Up to 4-5x cheaper per credit.

Yes. CRW is fully open source (AGPL-3.0). Docker setup takes under 5 minutes. Self-hosting is free.

Does CRW handle JavaScript-rendered pages?

Yes. Built-in browser rendering handles SPAs, React apps, and dynamic content. No Puppeteer or Playwright needed.

Each API call uses credits: /scrape = 1 credit, /crawl = 1 credit per page, /search = 1 credit per query. The free tier gives you 500 credits/month.

Is there an MCP server?

Yes. Native Model Context Protocol support. Connect CRW to Claude, Cursor, Windsurf, or any MCP-compatible AI agent.

Does CRW support structured JSON extraction with a schema?

Yes. Send formats: ["json"] together with a jsonSchema in your scrape request and CRW returns validated structured data using LLM function calling. Works with Anthropic and OpenAI models — compatible with Firecrawl v2 extraction format.

[ 200 OK ]

[ SCRAPE ]

[ .JSON ]

[ .MD ]

500 Free Credits — Get Started

Give your AI agents
fresh, reliable web data

Search, scrape, and extract clean data from any URL.
Ship trustworthy AI products without building scraping infrastructure. Open source & Rust-powered.

2.3x faster search than Tavily 92% coverage on 1K URLs 5.5x faster than Firecrawl

https://stripe.com/docs/api

[ .JSON ]

{
  "success": true,
  "data": {
    "url": "https://stripe.com/docs/api",
    "markdown": "# Stripe API Reference\n\nThe Stripe API is organized around REST...",
    "metadata": {
      "title": "Stripe API Reference",
      "ogTitle": "Stripe API Reference",
      "language": "en"
    },
    "links": [
      "https://stripe.com/docs/api/authentication",
      "https://stripe.com/docs/api/charges",
      "https://stripe.com/docs/api/customers"
    ]
  }
}

Scraped in 813ms

Integrations
Works with your stack

Python

JavaScript

Rust

Ruby

cURL

LangChain

CrewAI

n8n

Vercel

Dify

Botpress

Python

JavaScript

Rust

Ruby

cURL

LangChain

CrewAI

n8n

Vercel

Dify

Botpress

Python

JavaScript

Rust

Ruby

cURL

LangChain

CrewAI

n8n

Vercel

Dify

Botpress

[ 01 / 08 ] · MAIN FEATURES

Developer First

Search, scrape, crawl, and extract — one API

Four endpoints that cover the full pipeline. No glue code, no separate tools.

Search

Search the web and get full content from results.

Scrape

Get LLM-ready data from websites. Markdown, JSON, screenshot, etc.

Crawl

Crawl entire sites. All pages structured and returned as clean data.

Map

Discover all URLs on a domain. Fast sitemap generation.

1from crw import CRW
2
3crw = CRW(api_key="your-api-key")
4
5# Scrape any URL to clean markdown
6result = crw.scrape("https://example.com")
7print(result["markdown"])
8
9# Search the web with full page content
10results = crw.search("best rust web scrapers")
11for r in results:
12    print(r["title"], r["url"])

[ .MD ]

1# CRW
2
3CRW is a blazing-fast web scraping
4tool built in Rust. Extract clean
5data from any website.
6
7## Features
8
9- Scrape: Markdown from any page
10- Search: Search + scrape the web
11- Map: Discover all site URLs
12- Crawl: Full site extraction

/scrape~140ms

Scrape

Any URL to clean markdown, JSON, or screenshots. Full JS rendering without Puppeteer overhead.

/search~200ms

Search

Web search with full page content. Multi-engine aggregation with structured results.

/crawl~2s/page

Crawl

Entire sites, respect robots.txt. All pages structured and returned as clean data.

/map~500ms

Map

Discover all URLs on a domain. Fast sitemap generation without full page loads.

[ 02 / 08 ] · POWER YOUR AGENT

Integrations

Works where your
agents already live

Python, Node.js, cURL, MCP — drop into your existing stack in minutes.

One command

Skill. Give your agent easy access to real-time web data.

1npx crw-mcp

Quick config

MCP. Connect any MCP-compatible client to the web in seconds.

2 "mcpServers": {

3 "crw": { "command": "npx", "args": ["crw-mcp"] }

4 }

For AI agents

Agent Onboarding. Are you an AI agent? Fetch this skill to sign up, get an API key, and start building.

View the skill

MCP Compatible

Claude CodeClaude DesktopCursorWindsurfClineContinue.devOpenAI Codex CLI

Python SDK

crw on PyPI. Scrape, crawl, map with zero-config subprocess or cloud mode.

npm / npx

crw-mcp on npm. Cross-platform binary distribution for MCP and CLI.

LangChain

langchain-crw on PyPI. Document loader for RAG chains.

CrewAI

crewai-crw on PyPI. Full tool suite for CrewAI agents.

n8n

n8n-nodes-crw on npm. Visual workflow automation node.

Dify

crw-dify-plugin. Scrape, crawl, and map tools for Dify workflows.

[ 03 / 08 ] · CORE

Built for Performance

Benchmark-proven speed.
Open source transparency

Real benchmarks against production APIs. Every line of code is public.

No proxy headaches

Industry-leading reliability. Handles JS-heavy pages, anti-bot protections, and dynamic content. No proxies, no puppets, just clean data.

See benchmarks

Speed that feels invisible

Blazingly fast. Real benchmarks, not marketing math. Tested on 1,000+ URLs and queries against production APIs.

See full benchmark report

Scrapevs Firecrawl

2.2x faster

Searchvs Tavily

2.3x faster

Scrape Benchmark

1,000 URLs tested against production APIs

vs Firecrawl

Avg Latency1.5x

CRW

595ms

Firecrawl

866ms

Median Latency2.2x

CRW

255ms

Firecrawl

557ms

P95 Latency1.6x

CRW

1,236ms

Firecrawl

1,999ms

Search Benchmark

1,000 queries tested against production APIs

vs Tavily

Avg Latency2.3x

CRW

880ms

Tavily

2,000ms

Median Latency2.2x

CRW

785ms

Tavily

1,724ms

P95 Latency2.5x

CRW

1,433ms

Tavily

3,534ms

us/crwPublic

docs: restore Firecrawl/Tavily/Crawl4AI comparison table in Why CRW section

ef0c20f · Apr 13, 2026 · us

docs: remove star button from README body

f9f1e07 · Apr 12, 2026 · us

fix(ci): trigger release workflow after release-please creates tag

27f2b67 · Apr 12, 2026 · us

Integrations

Use well-known tools

Already fully integrated with the greatest existing tools and workflows.

See all integrations

Open Source

Code you can trust

Developed transparently and collaboratively. Join our community of contributors.

Check out our repo

[ 04 / 08 ] · FEATURES

Zero configuration

Ship faster without building infrastructure

JS rendering, anti-detection, media parsing, smart caching — handled. You focus on the product.

Docs to data

Media parsing. CRW can parse and output content from web hosted PDFs, DOCX, and more.

Knows the moment

Smart wait. CRW intelligently waits for content to load, making scraping faster and more reliable.

Selective persistence

Cached, when you need it. Selective caching — you choose your caching patterns, growing web index.

Advanced web coverage

Enhanced mode. Reaches every corner of the web with comprehensive coverage and high reliability.

Interactive scraping

Actions. Click, scroll, write, wait, press and more before extracting content.

Built-in stealth

Anti-detection. Rotates headers, handles CAPTCHAs, and mimics real browsers. No proxy setup needed.

[ 05 / 08 ] · PRICING

Transparent

Predictable pricing

Credit-based pricing with no surprises. Start with 500 free credits, scale as you grow.

MonthlyAnnual Save up to 15%

Free

$0/mo

500 credits

A lightweight way to try scraping.

Get Started

500 credits/month
2 concurrent requests
Community support

Hobby

$13$19/mo

3,000 credits

Great for side projects and small tools.

Launch pricing — ends June 1

Get Started

3,000 credits/month
5 concurrent requests
Basic support
+1,000 for $9

Popular

Standard

$69$99/mo

100,000 credits

Perfect for scaling with less effort.

Launch pricing — ends June 1

Get Started

100,000 credits/month
50 concurrent requests
Standard support
+35,000 for $47

Growth

$279$399/mo

500,000 credits

Built for high volume and speed.

Launch pricing — ends June 1

Get Started

500,000 credits/month
100 concurrent requests
Priority support
+175,000 for $177

Scale

$549$749/mo

1,000,000 credits

For teams scaling their data pipelines.

Launch pricing — ends June 1

Get Started

1,000,000 credits/month
150 concurrent requests
Priority support
+500,000 for $299

[ 06 / 08 ] · USE CASES

Use Cases

Built for teams shipping
production AI

See how teams use CRW for agents, RAG pipelines, research, and enrichment.

AI Chat & RAG

Feed clean web data to LLMs. Build RAG pipelines with structured markdown output.

Learn more

Lead Enrichment

Scrape company pages, LinkedIn profiles, and directories to enrich your CRM data.

Learn more

Market Research

Monitor competitors, track pricing changes, and analyze market trends at scale.

Learn more

AI Agents

Give your AI agents web access via MCP. Let them search, scrape, and interact autonomously.

Learn more

Content Aggregation

Crawl news sites, blogs, and forums. Aggregate content for analysis or republishing.

Learn more

Deep Research

Systematic web research with full-page extraction. Build knowledge bases from the open web.

Learn more

[ 07 / 08 ] · ENABLES

What CRW enables

Stop building scrapers.
Start shipping products.

Your agent needs live web data. CRW handles the hard part so you can focus on what matters.

Your agent needs fresh answers, not stale embeddings

Search + scrape live web results in one API call. Your agent always works with current information, not last month's crawl.

Stop assembling browser fleets to get clean data

Single binary, 6.6MB RAM. No Docker, no Puppeteer, no proxy rotation. Just deploy and call the API.

Drop into your agent as an MCP tool

Expose scrape, crawl, map, and search as MCP tools. Claude Code, Cursor, and custom agents get web access in one config.

Embed directly in your stack — no sidecar needed

Self-host the binary alongside your app. Same machine, same network. No external calls when latency matters.

Ship RAG that survives real users

Clean markdown from JS-heavy pages, SPAs, and anti-bot sites. Your pipeline gets structured content, not broken HTML.

Firecrawl-compatible API — switch your base URL

Same endpoint shapes, same SDK patterns. Migrate without rewriting a single line of integration code.

Know where every answer came from

Every response includes source URLs and metadata. Your users can verify, your team can debug, trust stays intact.

Your agent needs fresh answers, not stale embeddings

Search + scrape live web results in one API call. Your agent always works with current information, not last month's crawl.

Stop assembling browser fleets to get clean data

Single binary, 6.6MB RAM. No Docker, no Puppeteer, no proxy rotation. Just deploy and call the API.

Drop into your agent as an MCP tool

Expose scrape, crawl, map, and search as MCP tools. Claude Code, Cursor, and custom agents get web access in one config.

Embed directly in your stack — no sidecar needed

Self-host the binary alongside your app. Same machine, same network. No external calls when latency matters.

Ship RAG that survives real users

Clean markdown from JS-heavy pages, SPAs, and anti-bot sites. Your pipeline gets structured content, not broken HTML.

Firecrawl-compatible API — switch your base URL

Same endpoint shapes, same SDK patterns. Migrate without rewriting a single line of integration code.

Know where every answer came from

Every response includes source URLs and metadata. Your users can verify, your team can debug, trust stays intact.

Your agent needs fresh answers, not stale embeddings

Search + scrape live web results in one API call. Your agent always works with current information, not last month's crawl.

Stop assembling browser fleets to get clean data

Single binary, 6.6MB RAM. No Docker, no Puppeteer, no proxy rotation. Just deploy and call the API.

Drop into your agent as an MCP tool

Expose scrape, crawl, map, and search as MCP tools. Claude Code, Cursor, and custom agents get web access in one config.

Embed directly in your stack — no sidecar needed

Self-host the binary alongside your app. Same machine, same network. No external calls when latency matters.

Ship RAG that survives real users

Clean markdown from JS-heavy pages, SPAs, and anti-bot sites. Your pipeline gets structured content, not broken HTML.

Firecrawl-compatible API — switch your base URL

Same endpoint shapes, same SDK patterns. Migrate without rewriting a single line of integration code.

Know where every answer came from

Every response includes source URLs and metadata. Your users can verify, your team can debug, trust stays intact.

2.2x faster than Firecrawl, benchmarked in public

255ms median vs 557ms across 1,000 production URLs. All data, methodology, and code are public.

Search 2.3x faster than Tavily

785ms median vs 1,724ms across 1,000 queries. Real production APIs, not synthetic benchmarks.

Open source, self-host with zero license cost

AGPL-3.0. Run on your infra with unlimited requests. Audit every line. No vendor lock-in.

Rust core — no GC pauses, no runtime bloat

Memory-safe engine built for predictable latency. Your agents get consistent response times, not GC spikes.

5 endpoints cover the whole workflow

Scrape, crawl, map, search, and extract structured data. One tool instead of five different services.

Works with Python, Node.js, LangChain, n8n, and more

Official SDKs and native integrations. Add web data to your existing stack in minutes, not days.

Start free. Pay only when you scale.

500 free credits, no card required. Standard plan: $69/mo for 100K credits. Predictable, no surprise bills.

2.2x faster than Firecrawl, benchmarked in public

255ms median vs 557ms across 1,000 production URLs. All data, methodology, and code are public.

Search 2.3x faster than Tavily

785ms median vs 1,724ms across 1,000 queries. Real production APIs, not synthetic benchmarks.

Open source, self-host with zero license cost

AGPL-3.0. Run on your infra with unlimited requests. Audit every line. No vendor lock-in.

Rust core — no GC pauses, no runtime bloat

Memory-safe engine built for predictable latency. Your agents get consistent response times, not GC spikes.

5 endpoints cover the whole workflow

Scrape, crawl, map, search, and extract structured data. One tool instead of five different services.

Works with Python, Node.js, LangChain, n8n, and more

Official SDKs and native integrations. Add web data to your existing stack in minutes, not days.

Start free. Pay only when you scale.

500 free credits, no card required. Standard plan: $69/mo for 100K credits. Predictable, no surprise bills.

2.2x faster than Firecrawl, benchmarked in public

255ms median vs 557ms across 1,000 production URLs. All data, methodology, and code are public.

Search 2.3x faster than Tavily

785ms median vs 1,724ms across 1,000 queries. Real production APIs, not synthetic benchmarks.

Open source, self-host with zero license cost

AGPL-3.0. Run on your infra with unlimited requests. Audit every line. No vendor lock-in.

Rust core — no GC pauses, no runtime bloat

Memory-safe engine built for predictable latency. Your agents get consistent response times, not GC spikes.

5 endpoints cover the whole workflow

Scrape, crawl, map, search, and extract structured data. One tool instead of five different services.

Works with Python, Node.js, LangChain, n8n, and more

Official SDKs and native integrations. Add web data to your existing stack in minutes, not days.

Start free. Pay only when you scale.

500 free credits, no card required. Standard plan: $69/mo for 100K credits. Predictable, no surprise bills.

[ 08 / 08 ] · FAQ

FAQ

Frequently
asked questions

Everything you need to know about CRW.

Get started

Ready to ship?

Give your agent reliable web access in minutes. 500 free credits, no credit card required.

Get started — 500 free credits See pricing

Are you an AI agent? Get an API key here

Give your AI agentsfresh, reliable web data

Search, scrape, crawl, and extract — one API

Search

Scrape

Crawl

Map

Scrape

Search

Crawl

Map

Works where your agents already live

Skill. Give your agent easy access to real-time web data.

MCP. Connect any MCP-compatible client to the web in seconds.

Agent Onboarding. Are you an AI agent? Fetch this skill to sign up, get an API key, and start building.

Python SDK

npm / npx

LangChain

CrewAI

n8n

Dify

Benchmark-proven speed. Open source transparency

Industry-leading reliability. Handles JS-heavy pages, anti-bot protections, and dynamic content. No proxies, no puppets, just clean data.

Blazingly fast. Real benchmarks, not marketing math. Tested on 1,000+ URLs and queries against production APIs.

Use well-known tools

Code you can trust

Ship faster without building infrastructure

Media parsing. CRW can parse and output content from web hosted PDFs, DOCX, and more.

Smart wait. CRW intelligently waits for content to load, making scraping faster and more reliable.

Cached, when you need it. Selective caching — you choose your caching patterns, growing web index.

Enhanced mode. Reaches every corner of the web with comprehensive coverage and high reliability.

Actions. Click, scroll, write, wait, press and more before extracting content.

Anti-detection. Rotates headers, handles CAPTCHAs, and mimics real browsers. No proxy setup needed.

Predictable pricing

Free

Hobby

Standard

Growth

Scale

Built for teams shipping production AI

AI Chat & RAG

Lead Enrichment

Market Research

AI Agents

Content Aggregation

Deep Research

Stop building scrapers.Start shipping products.

Frequently asked questions

How is CRW faster than Firecrawl?

How is CRW cheaper?

Is the output quality the same?

Can I migrate from Firecrawl?

Can I self-host CRW?

Does CRW handle JavaScript-rendered pages?

What are credits?

Is there an MCP server?

Does CRW support structured JSON extraction with a schema?

Ready to ship?

Give your AI agents
fresh, reliable web data

Works where your
agents already live

Benchmark-proven speed.
Open source transparency

Built for teams shipping
production AI

Stop building scrapers.
Start shipping products.

Frequently
asked questions