Firecrawl MCP — Сценарии использования, Установка & Живое демо

Name: Firecrawl MCP Server
Author: firecrawl

Зачем использовать

Ключевые функции

Renders JavaScript — works on Next.js, React, Vue SPAs
Returns clean Markdown by default; raw HTML or structured JSON if you ask
Crawl entire sites with depth/path filters and concurrency control
Structured extraction with a JSON schema — no parsing prompts needed

Живое демо

Как выглядит на практике

firecrawl.replay ▶ готово

0/0

Установка

Выберите клиент

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ]
    }
  }
}

Откройте Claude Desktop → Settings → Developer → Edit Config. Перезапустите после сохранения.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ]
    }
  }
}

Cursor использует ту же схему mcpServers, что и Claude Desktop. Конфиг проекта приоритетнее глобального.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ]
    }
  }
}

Щёлкните значок MCP Servers на боковой панели Cline, затем "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ]
    }
  }
}

Тот же формат, что и Claude Desktop. Перезапустите Windsurf для применения.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "firecrawl",
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ]
    }
  ]
}

Continue использует массив объектов серверов, а не map.

~/.config/zed/settings.json

{
  "context_servers": {
    "firecrawl": {
      "command": {
        "path": "npx",
        "args": [
          "-y",
          "firecrawl-mcp"
        ]
      }
    }
  }
}

Добавьте в context_servers. Zed перезагружается автоматически.

claude mcp add firecrawl -- npx -y firecrawl-mcp

Однострочная команда. Проверить: claude mcp list. Удалить: claude mcp remove.

Сценарии использования

Реальные сценарии: Firecrawl

Build a live competitor pricing comparison table

👤 PMs, founders doing market research ⏱ ~20 min intermediate

Когда использовать: You're refreshing your pricing strategy and need to see what 5 competitors charge for comparable plans, with sources.

Предварительные требования

Firecrawl API key — firecrawl.dev — free tier gives 500 credits, enough for ~50 pages

Поток

Define a schema for what 'pricing' means

I want to compare pricing for [your category]. Define a schema that captures: plan name, monthly price, included usage, key features, and any per-unit overage costs.✓ Скопировано

→ Schema with these fields as strict types
Run extraction across 5 competitor URLs in parallel

Extract pricing from these 5 URLs using the schema: [list URLs]. Use the structured extraction endpoint.✓ Скопировано

→ 5 normalized JSON blobs you can compare
Render as a comparison table with citations

Render this as a Markdown table with each competitor as a column. Add a citation footnote pointing to each source URL.✓ Скопировано

→ Comparison ready to drop into a doc or deck

Итог: A normalized pricing table you can re-run anytime to detect changes — costs ~5 Firecrawl credits per refresh.

Подводные камни

Pricing pages often have toggles (monthly/annual, usage tiers) that scraping doesn't reveal — Run with actions: [{type: 'click', selector: '...toggle...'}] to capture both states, or note the limitation in your final output
Some sites block scrapers via Cloudflare — Firecrawl handles most. If still blocked, the page is likely using anti-bot measures intentionally — respect that and cite manually

Сочетать с: notion · google-sheets

Monitor a competitor's docs site for new features

👤 Product marketers, competitive intelligence ⏱ ~15 min intermediate

Когда использовать: You want to know within 24 hours when a competitor ships something new. Their changelog/blog is where they announce.

Предварительные требования

URL of their changelog or blog index page — Usually /changelog, /blog, /releases, or /whats-new

Поток

Crawl the changelog index

Crawl <URL> with depth=1, return all post URLs and their dates. Limit to 50 most recent.✓ Скопировано

→ List of recent posts with dates
Extract each post into a normalized format

For posts published in the last 30 days, extract title, date, summary, and any feature-name mentions, into JSON.✓ Скопировано

→ Structured changelog you can diff against last week's run
Highlight what's new vs last week

Compare to the previous run [paste prior JSON]. What's new this week and what's the strategic implication?✓ Скопировано

→ Diff with strategic commentary

Итог: A weekly competitor digest you can run as a scheduled job — never miss a feature drop.

Подводные камни

RSS feeds usually exist and are cheaper than scraping — Try <url>/feed.xml or <url>/rss first — only fall back to scraping if no feed

Сочетать с: notion

Enrich a list of company URLs with structured firmographic data

👤 Sales, marketing ops ⏱ ~30 min intermediate

Когда использовать: You have a CSV of 200 company websites and need their employee count, industry, and tech stack to prioritize outreach.

Предварительные требования

Source list of URLs — CSV or text file of homepage URLs

Поток

Define enrichment schema

Define a schema for: company_name, industry, employee_count_estimate, headquarters_location, primary_product_description (1 line), tech_stack_signals (list).✓ Скопировано

→ Strict JSON schema
Extract from each URL with rate limiting

For each URL in [list], run Firecrawl extract with the schema. Process in batches of 10 with 2-second pauses to stay polite.✓ Скопировано

→ Per-URL JSON, errors logged separately
Output a clean CSV

Convert results to CSV with the original URLs as the join key. Flag rows where extraction failed so I can retry manually.✓ Скопировано

→ CSV ready to import into your CRM

Итог: A 200-row enriched lead list in ~10 minutes for ~$2 in Firecrawl credits.

Подводные камни

Some companies just have a one-page splash with no info — Have Claude flag those for manual review rather than hallucinating
Cost adds up at scale — 10k URLs = real money — Sample first to validate schema quality before running on full list

Сочетать с: filesystem

Комбинации

Сочетайте с другими MCP — эффект x10

firecrawl + notion

Scrape, summarize, post directly to a Notion knowledge base

Scrape every post from anthropic.com/news in the last 30 days, summarize each in 2 sentences, and create a Notion page in 'AI News' database for each.✓ Скопировано

firecrawl + postgres

Scrape competitor data weekly and store in your DB for time-series analysis

Run my competitor-pricing extraction, then INSERT each row into the competitor_pricing table with today's date.✓ Скопировано

firecrawl + filesystem

Scrape docs sites and save as Markdown for offline RAG indexing

Crawl docs.stripe.com, save each page as a .md file under /knowledge/stripe/.✓ Скопировано

Инструменты

Что предоставляет этот MCP

Инструмент	Входные данные	Когда вызывать	Стоимость
firecrawl_scrape	url: str, formats?: ['markdown'\|'html'\|'links'\|'screenshot']	You need a single page's content as Markdown for the model	1 credit per page
firecrawl_crawl	url, limit, includePaths?, excludePaths?, maxDepth?	You need to scrape a whole section of a site	1 credit per page crawled
firecrawl_extract	urls: str[], schema: JSONSchema, prompt?	You want structured JSON, not raw text	5 credits per page (LLM-backed)
firecrawl_map	url, search?	Discover all URLs on a site without fetching content	1 credit

Стоимость и лимиты

Во что обходится

Квота API: Free tier: 500 credits, 10 concurrent. Hobby: $19/mo for 3000 credits.
Токенов на вызов: ~2000 tokens average per scraped page (Markdown body)
Деньги: 1 credit per scrape, 5 credits per structured extract, 1 per map. Credits don't roll over.
Совет: Use firecrawl_map first to plan, then only firecrawl_scrape the URLs you actually need. Cache aggressively — most pages don't change daily.

Безопасность

Права, секреты, радиус поражения

Хранение учётных данных: API key in env var FIRECRAWL_API_KEY

Исходящий трафик: All requests go via Firecrawl's servers (US/EU regions). They proxy out to target sites.

Firecrawl charges per page scraped — set crawl limits before unleashing on large sites.
Respect robots.txt and the site's ToS.

Устранение неполадок

Частые ошибки и исправления

402 Payment Required

Out of credits. Top up at firecrawl.dev/dashboard or wait for monthly reset.

Empty markdown returned

Page may be JS-only; ensure formats: ['markdown'] is set and mobile: false. Try with actions: [{type:'wait', milliseconds: 3000}] to allow JS execution.

Crawl returns fewer pages than expected

Check excludePaths isn't too aggressive. Also: many sites have a robots.txt blocking crawlers. Firecrawl honors it by default.

429 Too Many Requests (target site)

Reduce concurrency in your crawl options. Some sites enforce request budgets per IP.

Альтернативы

Firecrawl в сравнении

Альтернатива	Когда использовать	Компромисс
fetch MCP	Static HTML pages, no JS rendering needed, no schema extraction	Free and fast but breaks on SPAs and gives you raw HTML to parse
Playwright MCP	You need to interact with the page (login, click through flows)	More code, better for complex interactions; not optimized for bulk extraction
Brightdata MCP	You need residential proxies for sites that block datacenter IPs	Heavyweight and expensive — only worth it for tough targets
Apify MCP	You need pre-built scrapers for specific sites (Amazon, LinkedIn structured data)	Better for niche targets with prebuilt actors; more setup

Ещё

Ресурсы

📖 Читать официальный README на GitHub

🐙 Открытые задачи

🔍 Все 400+ MCP-серверов и Skills