Skip to main content

Building an AI startup?

You might be eligible for our Startup Program. Get fully funded access to the infrastructure you’re reading about right now (up to $20K value).
Bright Data MCP (Model Context Protocol) is a server that enables LLMs, AI agents, and applications to access, discover, and extract web data in real-time.
It allows MCP clients like Claude Desktop, Cursor, and Windsurf to search the web, navigate websites, take actions, and retrieve data without getting blocked, making it perfect for web scraping tasks.\
  • New MCP users get a free tier with 5,000 requests per month
The MCP server provides advanced capabilities, including:
  • Bypassing geo-restrictions to access content regardless of location
  • Navigating websites with bot detection protection using unlocker technology
  • Structured data extraction from platforms like Amazon, LinkedIn, Instagram, and many other data feeds
  • Remote browser automation for complex web interactions
  • Access to a global IP network to avoid blocking or rate limiting
Get started with Bright Data MCP:
  1. Explore the MCP GitHub repository to review the source code, contribute, or deploy your own instance.
  2. Read the MCP documentation to understand the API, request structure, and supported features.
  3. Get your API key
  4. Try the Bright Data AI Playground to test and experiment with live requests in a no-code environment.
Yes, the free tier includes up to 5,000 requests per month for web search and web page scraping via Web Unlocker.
You can get your API token from the user settings page in your Bright Data account. Make sure you have an account on brightdata.com - new users get free credit for testing, and pay-as-you-go options are available after that.
Bright Data MCP (Model Context Protocol) is a server that allows AI agents, LLMs, and dev tools to access and extract real-time web data using Bright Data’s infrastructure. It is available as both a remote (hosted) and local (self-hosted) server.To get started: BrightData MCP OverviewHelpful links:
Bright Data MCP allows you to integrate real-time web access into your AI agents or dev tools. It supports:Search & Scraping:
  • search_engine: Search Google, Bing, Yandex
  • scrape_as_markdown, scrape_as_html: Get content in clean formats
  • session_stats: Monitor tool usage
Structured Data Extraction:
  • Amazon: web_data_amazon_product, web_data_amazon_product_reviews
  • LinkedIn: web_data_linkedin_person_profile, web_data_linkedin_company_profile
  • Instagram: web_data_instagram_profiles, web_data_instagram_posts, web_data_instagram_reels
  • Facebook: web_data_facebook_posts, web_data_facebook_marketplace_listings
  • Others: web_data_x_posts, web_data_youtube_videos, web_data_zillow_properties_listing
Browser Automation:
  • scraping_browser_navigate, scraping_browser_click, scraping_browser_type
  • scraping_browser_get_html, scraping_browser_screenshot, scraping_browser_wait_for
🔗 Full list of tools: MCP Tools on GitHub
Bright Data MCP is designed to work with a wide range of AI tools and agents, including:
  • Claude Desktop (Anthropic)
  • Cursor IDE
  • Windsurf
  • LangChain
  • AutoGPT / AgentGPT
  • Smithery
  • Custom LLMs or AI agents that support MCP
These tools can use MCP to:
  • Perform real-time web searches
  • Scrape structured data
  • Automate browser interactions
  • Access geo-restricted or protected content
Try it live: Bright Data AI Playground
All you need is your Bright Data API key — the MCP server automatically creates and configures the required browser zone for you.To access browser tools, enable them using one of these methods:
  • Pro mode: Set &pro=1 (remote) or PRO_MODE=true (self-hosted) to enable all tools, including browser automation
  • Tool groups: Set &groups=browser (remote) or GROUPS=browser (self-hosted) to enable just the browser automation tools
For example, with the remote MCP:
https://mcp.brightdata.com/sse?token=YOUR_API_TOKEN&groups=browser
See the Remote Advanced Configuration or Local Advanced Configuration for more details.
For the full list of tools, please refer to our GitHub repo: MCP Tools.Bright Data MCP offers a comprehensive set of tools, including:Search and basic scraping:
  • search_engine: Scrape search results from Google, Bing, or Yandex
  • scrape_as_markdown: Scrape a webpage and get results in Markdown format
  • scrape_as_html: Scrape a webpage and get results in HTML format
  • session_stats: View tool usage during the current session
Structured data extraction:
  • Amazon: web_data_amazon_product, web_data_amazon_product_reviews
  • LinkedIn: web_data_linkedin_person_profile, web_data_linkedin_company_profile
  • ZoomInfo: web_data_zoominfo_company_profile
  • Instagram: web_data_instagram_profiles, web_data_instagram_posts, web_data_instagram_reels, web_data_instagram_comments
  • Facebook: web_data_facebook_posts, web_data_facebook_marketplace_listings, web_data_facebook_company_reviews
  • Various others: web_data_x_posts, web_data_zillow_properties_listing, web_data_booking_hotel_listings, web_data_youtube_videos
Browser automation:
  • scraping_browser_navigate: Navigate to a URL
  • scraping_browser_go_back/scraping_browser_go_forward: Navigate browser history
  • scraping_browser_click: Click on an element
  • scraping_browser_links: Get all links on the current page
  • scraping_browser_type: Type text into an element
  • scraping_browser_wait_for: Wait for an element to appear
  • scraping_browser_screenshot: Take a screenshot
  • scraping_browser_get_html/scraping_browser_get_text: Get page content
Some web data tools can take longer to execute, especially when dealing with complex websites. To ensure your agent can consume the data:
  1. Set a high enough timeout in your agent settings (180 seconds is recommended)
  2. For particularly slow sites, you may need to increase this value further
  3. Use the specialized web_data_* tools when available, as they’re often faster than general scraping
  4. For browser automation sequences, keep operations close together in time
Always treat scraped web content as untrusted data. Never use raw scraped content directly in LLM prompts to avoid potential prompt injection risks. Instead:
  • Filter and validate all web data before passing it to the LLM
  • Use structured data extraction rather than raw text when possible
  • Be cautious with executing JavaScript from scraped content
This error occurs when your system cannot find the npx command. To fix it:
  1. Find your Node.js path:
    • On macOS: Run which node in Terminal
    • On Windows: Run where node in Command Prompt
  2. Update your MCP configuration to use the full path instead of npx:
{
  "mcpServers": {
    "Bright Data": {
      "command": "/usr/local/bin/node", // Replace with your actual Node.js path
      "args": ["node_modules/@brightdata/mcp/index.js"],
      "env": {
        "API_TOKEN": "<your-api-token>"
      }
    }
  }
}
You can use the session_stats tool to check your usage during the current session. For comprehensive usage tracking and billing information, log in to your Bright Data account dashboard. The session_stats tool will show you information about:
  • Number of requests made during the current session
  • Tools used and their frequency
Yes, you can try Bright Data MCP without any setup using the Bright Data AI Playground. It provides an easy way to explore the capabilities of Bright Data MCP without any local setup. Just sign in and start experimenting with web data collection!
You can connect your AI agent with Bright Data by following these steps:
  1. Install Node.js (only required for the self-hosted option)
  2. Create a Bright Data account (new users get free credit for testing)
  3. Copy your API key from the user settings page (new users receive an API key in the welcome email)
  4. Configure your agent with your API key. For more detailed examples, visit us on GitHub or check our MCP documentation.
No, zones are created automatically for you. All you need is Node.js (for self-hosted) and your Bright Data API key. If you want to change your zone name to something specific, you can always do so through the control panel.
Yes! Bright Data provides a free tier of 5,000 requests/month to new MCP users. The free tier includes web search and web page scraping via Web Unlocker. See What is the MCP free tier? below for more details.
The MCP free tier allows you to connect your AI agent via MCP (Model Context Protocol) to Bright Data services for free. You get up to 5,000 requests per month for web search and scraping any public webpage via Web Unlocker at no cost. This is ideal for testing, development, or light usage.For more information, see the Remote MCP Quickstart.
When you exceed the 5,000 monthly request limit on the free tier, requests will start failing. To continue using the MCP server, you need to create a new Web Unlocker zone and configure the MCP to use it:
  1. Go to brightdata.com/cp and click Add (top-right)
  2. Select Unlocker zone and create it with any name (e.g., my_unlocker)
  3. Update your MCP configuration to use the new zone:
    • Remote MCP: Add &unlocker=ZONE_NAME to your MCP URL
    • Self-hosted MCP: Set the WEB_UNLOCKER_ZONE=ZONE_NAME environment variable
  4. Restart your MCP client (e.g., Claude Desktop, Cursor) after the configuration change
Yes, you can run multiple MCP instances for parallelized workloads. However, if you are using the self-hosted (local) MCP with npx, each instance downloads and caches the package separately, which can consume significant disk space.For parallel use cases, we recommend using the Remote MCP instead. It requires no local installation, runs entirely in the cloud, and eliminates disk space concerns. Simply point each instance to the remote endpoint:
https://mcp.brightdata.com/sse?token=YOUR_API_TOKEN
See the Remote MCP Quickstart for setup instructions.