Web MCP

Web MCP

What is Web MCP about?

Web MCP enables LLMs and AI agents to query, navigate, and extract information from the live web without being blocked. It integrates seamlessly with any MCP‑compatible assistant, delivering up‑to‑date content, geo‑unrestricted access, and advanced tools such as browser control and structured data extraction.

How to use Web MCP?

1. Obtain an API token from a Bright Data account.
2. Start the server (e.g., npx @brightdata/mcp). The server listens for MCP requests at https://mcp.brightdata.com/mcp.
3. Configure your AI assistant by adding the remote MCP URL (including the token) in the connector or developer settings of the assistant (Claude Desktop, Cursor, etc.).
4. Send requests using the supported tools (search_engine, scrape_as_markdown, and, in Pro mode, the full suite of ~60 tools).

Key features of Web MCP

• Real‑time web search and data retrieval.
• Geo‑restriction bypass and anti‑bot detection handling.
• Web Unlocker for sites with bot protection.
• Remote browser automation (Pro mode).
• Structured data extraction tools (web_data_*).
• Rate limiting and zone customization via environment variables.
• Free tier: 5,000 requests per month for the first three months.

Use cases of Web MCP

• Dynamic information lookup for chat‑based agents (e.g., market caps, weather forecasts).
• Content gathering for research assistants that need the latest articles or product listings.
• Automation workflows that require navigating login‑protected or captcha‑protected sites.
• Data enrichment where structured web data is merged with internal databases.
• Prototyping AI‑driven agents with live web interaction without managing proxy infrastructure.

FAQ

Q: Do I need to pay after the free tier? A: The free tier provides 5,000 requests per month for three months. After that, a credit card is required; charges only apply if you exceed the free quota or enable Pro‑mode tools.

Q: Which tools are available by default? A: By default only search_engine and scrape_as_markdown are exposed. Enabling PRO_MODE=true unlocks the full toolset, including browser automation and structured data extraction.

Q: How do I set rate limits? A: Define the RATE_LIMIT environment variable (e.g., RATE_LIMIT=100/1h). The format is <number>/<period><unit> where units are s, m, or h.

Q: Can I use a custom proxy zone? A: Yes. Set WEB_UNLOCKER_ZONE and BROWSER_ZONE to the names of zones you created in the Bright Data control panel.

Q: What environment variables are required? A: At minimum API_TOKEN (your Bright Data API key). Optional variables include PRO_MODE, RATE_LIMIT, WEB_UNLOCKER_ZONE, and BROWSER_ZONE.

🌟 Overview

Welcome to the official Bright Data's Web MCP server, solving web access for LLMs and AI agents by allowing them to effectively search, extract and navigate the web without getting blocked. The Web MCP supports all major LLMs, IDEs and agent frameworks (either locally hosted, SSE or Streamable HTTP), enabling your tools to seamlessly search the web, navigate websites, take action and retrieve data - without getting blocked.

🚀 The Web MCP includes 5,000 free requests each month - ideal for your everyday use and for prototyping smart agentic workflows.

Note: The Web MCP free tier offers 5,000 requests per month for the first 3 months. After that, a credit card will be required, but there will be no extra charges unless premium features like mcp_browser or Web Scrapers are used.

Table of Content

• 🎬 Demo
• ✨ Features
• 💡 Usage Examples
• 🚀 Quickstart with Claude Desktop
• 🔧 Available Tools
• ⚠️ Security Best Practices
• 🔧 Account Setup
• 🔌 Other MCP Clients
• 🎮 Try Bright Data MCP Playgrounds
• ⚠️ Troubleshooting
• 👨‍💻 Contributing
• 📞 Support

🎬 Demo

The videos below demonstrate a minimal use case for Claude Desktop:

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

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

For more YouTube tutorials and demos: Demo

✨ Features

• Real-time Web Access: Access up-to-date information directly from the web
• Bypass Geo-restrictions: Access content regardless of location constraints
• Web Unlocker: Navigate websites with bot detection protection
• Browser Control: Remote browser automation capabilities
• Seamless Integration: Works with all MCP-compatible AI assistants

💡 Usage Examples

Some example queries that this MCP server will be able to help with:

• "Google some movies that are releasing soon in [your area]"
• "What's Tesla's current market cap?"
• "What's the Wikipedia article of the day?"
• "What's the 7-day weather forecast in [your location]?"
• "Of the 3 highest paid tech CEOs, how long have their careers been?"

Quickstart with Cursor

🚀 Quickstart with hosted MCP on Claude Desktop

Through Connectors:

1. Open Claude Desktop

2. Go to: Settings → Connectors → Add custom connector

3. Choose a Name, and in the “Remote MCP server URL” section, paste:

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

4. Replace YOUR_API_TOKEN_HERE with your actual API token from Step 1, and click “Add”

Through Developer Settings:

1. Open Claude Desktop

2. Go to: Settings → Developer → Edit Config

3. Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.brightdata.com/mcp?token=YOUR_API_TOKEN_HERE"
      ]
    }
  }
}

4. Replace YOUR_API_TOKEN_HERE with your actual API token from Step 1

5. Save and restart Claude Desktop

💻 Use with self hosted MCP on Claude Desktop

Through Claude Desktop Extension:

1. Download the Claude Desktop Extension:
   📦 Bright Data's MCP Extension

2. Open Claude and go to:
   Settings → Extensions

3. Drag the .dtx file from Step 1 into the dropping area.

4. Enable the service and restart Claude.

5. Enjoy!

Through claude_desktop_config.json:

1. Install nodejs to get the npx command (node.js module runner). Installation instructions can be found on the node.js website

2. Go to Claude > Settings > Developer > Edit Config > Edit "claude_desktop_config.json" to include the following:

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

🛸 Or for more advanced options:

{
  "mcpServers": {
    "Bright Data": {
      "command": "npx",
      "args": ["@brightdata/mcp"],
      "env": {
        "API_TOKEN": "<insert-your-api-token-here>",
        "RATE_LIMIT": "<optional if you want to change rate limit format: limit/time+unit, e.g., 100/1h, 50/30m, 10/5s>",
        "WEB_UNLOCKER_ZONE": "<optional if you want to override the web unlocker zone name - default is mcp_unlocker>",
        "BROWSER_ZONE": "<optional if you want to override the browser zone name - defaults is mcp_browser>",
        "PRO_MODE": "<optional boolean, defaults to false. Set to true to expose all tools including browser and web_data_* tools>"
      }
    }
  }
}

🔧 Available Tools

Important: Pro mode is not included in the free tier and will incur additional charges. If you choose to use Pro mode, you’ll gain access to all 60 tools but please be aware of the associated costs.

To enable Pro mode, simply add "PRO_MODE"=true to your enviroment variables.

List of Available Tools

Note: By default, only basic tools (search_engine and scrape_as_markdown) are exposed. To access all tools including browser automation and web data extraction, enable PRO_MODE in your configuration (see Account Setup section).

⚠️ Security Best Practices

Important: 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 processing
• Use structured data extraction rather than raw text (web_data tools)

🔧 Account Setup

1. Make sure you have an account on brightdata.com (new users get free credit for testing, and pay as you go options are available)

2. Get your API key from the user settings page, or from the welcome email you received

Optional:

3. Enable Pro Mode (for access to all tools):

   • Set PRO_MODE=true in your environment configuration to access browser automation, structured data extraction, and all available tools
   • Default: false (only exposes search_engine and scrape_as_markdown tools)
   • See the advanced configuration example above for implementation details

4. Configure rate limiting:

   • Set the RATE_LIMIT environment variable to control API usage
   • Format: limit/time+unit (e.g., 100/1h for 100 calls per hour)
   • Supported time units: seconds (s), minutes (m), hours (h)
   • Examples: RATE_LIMIT=100/1h, RATE_LIMIT=50/30m, RATE_LIMIT=10/5s
   • Rate limiting is session-based (resets when server restarts)

5. Create a custom Web Unlocker zone

   • By default, we create a Web Unlocker zone automatically using your API token
   • For more control, you can create your own Web Unlocker zone in your control panel and specify it with the WEB_UNLOCKER_ZONE environment variable

6. Create a custom Browser API zone:

   • By default, we create a Browser API zone automatically using your API token.
   • For more control, you can create your own Browser API zone in your control panel and specify it with the BROWSER_ZONE environment variable

🔌 Other MCP Clients

To use this MCP server with other agent types, you should adapt the following to your specific software:

• Before running the server, make sure the API_TOKEN=<your-token> environment variable is set
• The full command to run the MCP server is npx @brightdata/mcp

💻 macOS / Linux (bash/zsh)

export API_TOKEN=your-token
npx @brightdata/mcp

🪟 Windows (Command Prompt)

set API_TOKEN=your-token
npx @brightdata/mcp

🪟 Windows (PowerShell)

$env:API_TOKEN="your-token"
npx @brightdata/mcp

💡 Tip: You can also use a .env file and a tool like dotenv to manage environment variables more easily during development.

---

🔄 Changelog

CHANGELOG.md

🎮 Try Bright Data MCP Playgrounds

Want to try Bright Data MCP without setting up anything?

Check out this playground on Smithery:

This platform 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!

⚠️ Troubleshooting

Timeouts when using certain tools

Some tools can involve reading web data, and the amount of time needed to load the page can vary by quite a lot in extreme circumstances.

To ensure that your agent will be able to consume the data, set a high enough timeout in your agent settings.

A value of 180s should be enough for 99% of requests, but some sites load slower than others, so tune this to your needs.

spawn npx ENOENT

This error occurs when your system cannot find the npx command. To fix it:

Finding npm/Node Path

macOS:

which node

Shows path like /usr/local/bin/node

Windows:

where node

Shows path like C:\Program Files\nodejs\node.exe

Update your MCP configuration:

Replace the npx command with the full path to Node, for example, on mac, it will look as follows:

"command": "/usr/local/bin/node"

👨‍💻 Contributing

We welcome contributions to help improve the Bright Data MCP! Here's how you can help:

1. Report Issues: If you encounter any bugs or have feature requests, please open an issue on our GitHub repository.
2. Submit Pull Requests: Feel free to fork the repository and submit pull requests with enhancements or bug fixes.
3. Coding Style: All JavaScript code should follow Bright Data's JavaScript coding conventions. This ensures consistency across the codebase.
4. Documentation: Improvements to documentation, including this README, are always appreciated.
5. Examples: Share your use cases by contributing examples to help other users.

For major changes, please open an issue first to discuss your proposed changes. This ensures your time is well spent and aligned with project goals.

📞 Support

If you encounter any issues or have questions, please reach out to the Bright Data support team or open an issue in the repository.