Bright Data MCP — Сценарии использования, Установка & Живое демо

Name: Bright Data MCP Server
Author: brightdata

Зачем использовать

Ключевые функции

SERP API — real-time Google/Bing results without getting blocked
Web Unlocker — fetch any public URL through rotating residential IPs
Structured scrapers for 50+ popular targets with normalized JSON output
Browser automation option for JS-heavy / multi-step flows
Bulk mode: send a list of URLs, get results back as a dataset

Живое демо

Как выглядит на практике

brightdata.replay ▶ готово

0/0

Установка

Выберите клиент

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

{
  "mcpServers": {
    "brightdata": {
      "command": "npx",
      "args": [
        "-y",
        "@brightdata/mcp"
      ]
    }
  }
}

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

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "brightdata": {
      "command": "npx",
      "args": [
        "-y",
        "@brightdata/mcp"
      ]
    }
  }
}

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

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "brightdata": {
      "command": "npx",
      "args": [
        "-y",
        "@brightdata/mcp"
      ]
    }
  }
}

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

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "brightdata": {
      "command": "npx",
      "args": [
        "-y",
        "@brightdata/mcp"
      ]
    }
  }
}

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

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "brightdata",
      "command": "npx",
      "args": [
        "-y",
        "@brightdata/mcp"
      ]
    }
  ]
}

Continue использует массив объектов серверов, а не map.

~/.config/zed/settings.json

{
  "context_servers": {
    "brightdata": {
      "command": {
        "path": "npx",
        "args": [
          "-y",
          "@brightdata/mcp"
        ]
      }
    }
  }
}

Добавьте в context_servers. Zed перезагружается автоматически.

claude mcp add brightdata -- npx -y @brightdata/mcp

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

Сценарии использования

Реальные сценарии: Bright Data

Track your keyword rankings on Google across locations

👤 SEO teams ⏱ ~20 min intermediate

Когда использовать: You want daily rank tracking for 50 keywords in US/UK/DE without running your own proxies.

Предварительные требования

Bright Data API token — brightdata.com → dashboard → API tokens
Budget: ~$0.001–$0.003 per SERP query — Credit balance on Bright Data account

Поток

Run the SERP for each kw/country

For each keyword in [list], run a Google SERP search from country=us. Capture top 10 organic results (url, title, position).✓ Скопировано

→ Per-keyword ranked list
Locate our domain

For each result set, find where mydomain.com appears (or 'not in top 10'). Output kw → position.✓ Скопировано

→ Rank table
Diff vs yesterday

Compare to yesterday's JSON [paste]. Highlight moves > 3 positions.✓ Скопировано

→ Daily movers report

Итог: A daily rank-tracking workflow at ~$0.15/day for 50 keywords, no proxy ops.

Подводные камни

Each country/device combo counts as a separate query — Only track what you need; 50 kw × 3 countries × 7 days is 1050 queries/week

Сочетать с: postgres · notion

Fetch a page that blocks datacenter IPs

👤 Devs whose normal scraping target got behind Cloudflare-Turnstile ⏱ ~15 min intermediate

Когда использовать: fetch / Firecrawl get 403 or an interstitial; you need residential IPs.

Поток

Try once with unlocker

Fetch <url> via Web Unlocker. Return the rendered HTML + HTTP status.✓ Скопировано

→ 200 + real HTML
Extract what you need

From that HTML, extract [list the fields]. Return as JSON.✓ Скопировано

→ Structured data
Respect the site

If the page says 'robots.txt disallow' or a clear anti-scrape notice, abort and tell me.✓ Скопировано

→ Consent-aware fallback

Итог: The data you need without maintaining a proxy pool.

Подводные камни

Unblocker can still fail on hardcore targets (banking, SaaS login pages) — These are intentionally private; pick an official API or a different approach
Costs escalate quickly on large crawls — Firecrawl or fetch is cheaper for unprotected sites — only pay Bright Data when you hit a block

Сочетать с: firecrawl

Pull a structured Amazon product dataset via prebuilt scraper

👤 E-commerce analysts ⏱ ~20 min intermediate

Когда использовать: You want 500 Amazon ASINs with clean title/price/rating/bsr fields, not raw HTML.

Предварительные требования

List of ASINs or category URLs — CSV or text input

Поток

Kick off the prebuilt Amazon scraper

Run the Bright Data Amazon product scraper for ASINs [list]. Return a job id.✓ Скопировано

→ Job id issued
Poll until ready

Poll the job. When done, fetch the dataset.✓ Скопировано

→ Full dataset delivered
Cache to avoid re-runs

Save the dataset to /data/amazon-<date>.jsonl. Flag any ASIN that errored.✓ Скопировано

→ Persisted dataset + error list

Итог: A clean, re-runnable Amazon product dataset at ~$X/1000 products (see current pricing).

Подводные камни

Public LinkedIn/Instagram scrapers legally vary by region — Stay within public profile data; do not bypass authentication — know your jurisdiction

Сочетать с: postgres · filesystem

Daily news search for a brand across SERP

👤 PR / comms ⏱ ~15 min beginner

Когда использовать: You want a daily digest of what's being said about your brand in news search.

Поток

Run a Google News SERP

Google News search for '<brand>' last 24h, country=us. Return top 20 results with source, title, url, snippet.✓ Скопировано

→ News SERP
Classify sentiment from snippets

Score each result as positive/neutral/negative based on title+snippet. Flag anything flagged as negative for review.✓ Скопировано

→ Scored list
Deliver the digest

Format as a markdown digest: counts by sentiment, negative items with links, top positive items.✓ Скопировано

→ Digest ready

Итог: A focused PR digest without scraping individual news sites.

Подводные камни

Sentiment from headlines alone is noisy — Only flag as negative if both title and snippet are clearly negative; human-review the flags

Сочетать с: notion

Комбинации

Сочетайте с другими MCP — эффект x10

brightdata + postgres

Store daily rank-tracking rows for trend analysis

Run SERP for kw list, INSERT into keyword_ranks table with today's date.✓ Скопировано

brightdata + firecrawl

Cheap-first, unblock-fallback crawling

Try Firecrawl first; if 403/blocked, fall back to Bright Data Unlocker for that URL only.✓ Скопировано

brightdata + notion

Weekly PR digest posted to Notion

Run brand SERP for the last 7 days, create a Notion page with the digest.✓ Скопировано

Инструменты

Что предоставляет этот MCP

Инструмент	Входные данные	Когда вызывать	Стоимость
search_engine	engine: 'google'\|'bing'\|'duckduckgo', query, country?, lang?, device?	SERP / rank-tracking workflows	~$0.001–0.003 per query
scrape_as_markdown	url	Fetch a single page through unblocker as clean markdown	1 credit per page
scrape_as_html	url	You need raw HTML to parse yourself	1 credit per page
web_data_<target>	urls: str[] or params	Prebuilt structured scraper (amazon, linkedin, zillow, etc.)	per-scraper pricing
scraping_browser_*	url, actions	Multi-step / JS-heavy flows	browser-session pricing

Стоимость и лимиты

Во что обходится

Квота API: Bounded by account credits; concurrent requests per plan
Токенов на вызов: SERP: 500–2000 tokens. Scraped page: 1000–5000 tokens.
Деньги: Pay-as-you-go; typical SERP $0.001–$0.003, unblocker ~$3 per 1000 pages, prebuilt scrapers priced per 1000 records.
Совет: Cache aggressively — most data doesn't change hourly. Use cheaper fetch/Firecrawl for unprotected targets.

Безопасность

Права, секреты, радиус поражения

Минимальные скоупы: API token with zone access for the relevant products

Хранение учётных данных: BRIGHTDATA_API_TOKEN in env

Исходящий трафик: All requests through Bright Data's proxy network; they see target URLs and responses

Никогда не давайте: Admin-level account tokens for everyday scraping

Respect robots.txt and the target site's ToS — Bright Data doesn't absolve you.
Scraping private/logged-in content is out of scope and may violate both BD ToS and the target's.

Устранение неполадок

Частые ошибки и исправления

401 Invalid token

BRIGHTDATA_API_TOKEN missing/expired. Regenerate in dashboard.

Проверить: curl -H 'Authorization: Bearer $BRIGHTDATA_API_TOKEN' https://api.brightdata.com/zone/list

402 Insufficient credits

Top up account balance or reduce query volume; check dashboard for burn rate.

Scraping job SUCCEEDED but dataset empty

Wrong input schema for the prebuilt scraper. Read the scraper's doc page for required fields.

Target site still blocks despite unlocker

Some sites use advanced fingerprinting; switch to Scraping Browser with stealth, or abandon the target.

Альтернативы

Bright Data в сравнении

Альтернатива	Когда использовать	Компромисс
Firecrawl MCP	Unprotected sites, generic scraping	Fails on hostile targets
Apify MCP	You want a broader Actor marketplace and cheaper pricing for common targets	Proxy network quality varies per Actor
SerpAPI MCP	You only need SERP, not full scraping	No unblocker / prebuilt scrapers

Ещё

Ресурсы

📖 Читать официальный README на GitHub

🐙 Открытые задачи

🔍 Все 400+ MCP-серверов и Skills