convez.ai
MCP ServerIntegração & APIsIntermediário
brightdata

Bright Data Web MCP

brightdata/brightdata-mcp

Superpoderes web em tempo real para LLMs, sem bloqueios

Ver no GitHub
BomPluginScore · de 0 a 100

Servidor MCP que conecta Claude e outros LLMs diretamente à web em tempo real, desbloqueando acesso a dados públicos sem ser bloqueado. Inclui ferramentas para web scraping, consulta a LLMs concorrentes (ChatGPT, Grok, Perplexity) e acesso a metadados de pacotes npm/PyPI para agentes de código.

Instalação

$npx convez-ai brightdata/brightdata-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
2,7 mil
Forks
336
Acompanhando
11
Saúde
37%
Último commit
há 1 dia
Versão
v4.8.0
Issues abertas
39
Contribuidores
16

Casos de uso

  • Pesquisa e coleta de dados em tempo real da web
  • Monitoramento de menções de marca em LLMs concorrentes (GEO)
  • Agentes de código consultando metadados de dependências npm/PyPI
  • Análise competitiva e visibilidade em mecanismos generativos
  • Prototipagem rápida com tier gratuito (5.000 créditos/mês)

Veredito

MCP maduro e bem mantido que resolve genuinamente o problema de acesso web para LLMs. A abordagem de grupos de ferramentas (web, code, geo) é inteligente e o tier gratuito remove fricção para experimentação. Recomendado para quem precisa de dados web em tempo real em workflows com Claude.

README

A tradução para o português deste README ainda está na fila.


🌟 Overview

The Web MCP is your gateway to giving AI assistants true web capabilities. No more outdated responses, no more "I can't access real-time information" - just seamless, reliable web access that actually works.

Built by Bright Data, the world's #1 web data platform, this MCP server ensures your AI never gets blocked, rate-limited, or served CAPTCHAs.


🎯 Perfect For

  • 🔍 Real-time Research - Get current prices, news, and live data
  • 🛍️ E-commerce Intelligence - Monitor products, prices, and availability
  • 📊 Market Analysis - Track competitors and industry trends
  • 🤖 AI Agents - Build agents that can actually browse the web
  • 💻 Coding Agents - Look up npm/PyPI packages, versions, and READMEs in real time
  • 🧠 GEO & Brand Visibility - See how ChatGPT, Grok, and Perplexity perceive your brand
  • 📝 Content Creation - Access up-to-date information for writing
  • 🎓 Academic Research - Gather data from multiple sources efficiently

⚡ Quick Start

Use the configuration wizard:

GIF for day2

Perfect for users who want zero setup. Just add this URL to your MCP client:

https://mcp.brightdata.com/mcp?token=YOUR_API_TOKEN_HERE

Setup in Claude Desktop:

  1. Go to: Settings → Connectors → Add custom connector
  2. Name: Bright Data Web
  3. URL: https://mcp.brightdata.com/mcp?token=YOUR_API_TOKEN
  4. Click "Add" and you're done! ✨
{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": ["@brightdata/mcp"],
      "env": {
        "API_TOKEN": "<your-api-token-here>"
      }
    }
  }
}

🚀 Pricing & Modes

💡 How the free tier works: Your account gets 5,000 credits/month (~$7.50 value), auto-renewed on the 1st with no credit card required. MCP requests draw from this shared pool — base tools (search_engine, scrape_as_markdown, discover) cost 1 credit per request, while web_data_* tools cost 1 credit per record returned. Unused credits don't roll over, and when they run out requests stop (no surprise charges unless you've deposited funds). Pro mode is not included in the free tier and incurs additional charges based on usage.


🧠 Advanced Tool Selection

  • GROUPS lets you enable curated tool bundles. Use comma-separated group IDs such as ecommerce,browser.
  • TOOLS adds explicit tool names on top of the selected groups.
  • Mode priority: PRO_MODE=true (all tools) → GROUPS / TOOLS (whitelist) → default rapid mode (base toolkit).
  • Base tools always enabled: search_engine, search_engine_batch, scrape_as_markdown, scrape_batch, discover.
  • Group ID custom is reserved; use TOOLS for bespoke picks.

Claude Desktop example

{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": ["@brightdata/mcp"],
      "env": {
        "API_TOKEN": "<your-api-token-here>",
        "GROUPS": "browser,advanced_scraping",
        "TOOLS": "extract"
      }
    }
  }
}

Coding agent example (Claude Code / Cursor / Windsurf)

Give your coding agent real-time package intelligence — latest versions, READMEs, dependencies, and metadata from npm and PyPI without scraping:

{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": ["@brightdata/mcp"],
      "env": {
        "API_TOKEN": "<your-api-token-here>",
        "GROUPS": "code"
      }
    }
  }
}

✨ Features

🔥 Core Capabilities

🎯 Example Queries That Just Work

✅ "What's Tesla's current stock price?"
✅ "Find the best-rated restaurants in Tokyo right now"
✅ "Get today's weather forecast for New York"
✅ "What movies are releasing this week?"
✅ "What are the trending topics on Twitter today?"
✅ "What's the latest version of express on npm?"
✅ "Get the README for the langchain-brightdata PyPI package"

🎬 Demos

Note: These videos show earlier versions. New demos coming soon! 🎥

Basic Web Search Demo

https://github.com/user-attachments/assets/59f6ebba-801a-49ab-8278-1b2120912e33

Advanced Scraping Demo

https://github.com/user-attachments/assets/61ab0bee-fdfa-4d50-b0de-5fab96b4b91d

📺 More tutorials on YouTube →


🔧 Available Tools

⚡ Rapid Mode Tools (Default - Free)

ToolDescriptionUse Case
🔍 search_engineWeb search with AI-optimized resultsResearch, fact-checking, current events
📄 scrape_as_markdownConvert any webpage to clean markdownContent extraction, documentation
🎯 discoverAI-ranked web search with intent-based relevance scoringDeep research, RAG pipelines, competitive analysis

💎 Pro Mode Tools (60+ Tools)

CategoryToolsDescription
Browser Controlscraping_browser.*Full browser automation
Web Data APIsweb_data_*Structured data extraction
E-commerceProduct scrapersAmazon, eBay, Walmart data
Social MediaSocial scrapersTwitter, LinkedIn, Instagram
Maps & LocalLocation toolsGoogle Maps, business data

📚 View complete tool documentation →


🎮 Try It Now!

🧪 Online Playground

Try the Web MCP without any setup:


🔧 Configuration

Basic Setup

{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": ["@brightdata/mcp"],
      "env": {
        "API_TOKEN": "your-token-here"
      }
    }
  }
}

Advanced Configuration

{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": ["@brightdata/mcp"],
      "env": {
        "API_TOKEN": "your-token-here",
        "PRO_MODE": "true",              // Enable all 60+ tools
        "RATE_LIMIT": "100/1h",          // Custom rate limiting
        "WEB_UNLOCKER_ZONE": "custom",   // Custom unlocker zone
        "BROWSER_ZONE": "custom_browser", // Custom browser zone
        "POLLING_TIMEOUT": "600"         // Polling timeout in seconds (default: 600)
      }
    }
  }
}

Environment Variables

VariableDescriptionDefaultExample
API_TOKENYour Bright Data API token (required)-your-token-here
PRO_MODEEnable all 60+ toolsfalsetrue
RATE_LIMITCustom rate limitingunlimited100/1h, 50/30m
WEB_UNLOCKER_ZONECustom Web Unlocker zone namemcp_unlockermy_custom_zone
BROWSER_ZONECustom Browser zone namemcp_browsermy_browser_zone
POLLING_TIMEOUTTimeout for web_data_* tools polling (seconds)600300, 1200
BASE_TIMEOUTRequest timeout for base tools in seconds (search & scrape)No limit60, 120
BASE_MAX_RETRIESMax retries for base tools on transient errors (0-3)01, 3
GROUPSComma-separated tool group IDs-ecommerce,browser
TOOLSComma-separated individual tool names-extract,scrape_as_html

Notes:

  • POLLING_TIMEOUT controls how long web_data_* tools wait for results. Each second = 1 polling attempt.
  • Lower values (e.g., 300) will fail faster on slow data collections.
  • Higher values (e.g., 1200) allow more time for complex scraping tasks.

📚 Documentation


🚨 Common Issues & Solutions

❌ "spawn npx ENOENT" Error

Solution: Install Node.js or use the full path to node:

"command": "/usr/local/bin/node"  // macOS/Linux
"command": "C:\\Program Files\\nodejs\\node.exe"  // Windows

⏱️ Timeouts on Complex Sites

Solution: Increase timeout in your client settings to 180s

🔑 Authentication Issues

Solution: Ensure your API token is valid and has proper permissions

📡 Remote Server Connection

Solution: Check your internet connection and firewall settings

More troubleshooting →


🤝 Contributing

We love contributions! Here's how you can help:

Please follow Bright Data's coding standards.


📞 Support


📜 License

MIT © Bright Data Ltd.


Ver o original no GitHub

Bom para

Relacionados