brightdata-mcp: Give AI Agents Access to Web Data

Overview

brightdata-mcp is a JavaScript Model Context Protocol server that gives compatible AI agents access to public web data. It combines search, page scraping, platform-specific structured data tools, and remote browser automation behind one MCP interface.

The project addresses a common integration problem: ordinary web requests can be blocked by bot detection, CAPTCHAs, rate limits, or geographic restrictions. Requests use Bright Data's infrastructure, so this can simplify collection workflows, but the service requires a Bright Data account and API token.

Key Features

Search Google, Bing, and Yandex, with batch search and relevance-ranked discovery options.
Scrape URLs as Markdown or HTML, and batch scrape up to 10 URLs per call.
Retrieve structured data from supported platforms, including e-commerce, social, business, and app-store sources.
Use AI-assisted extraction to turn arbitrary pages into structured JSON.
Automate a remote browser with navigation, clicks, typing, screenshots, and page snapshots.
Connect through a hosted MCP endpoint or run locally with npx @brightdata/mcp.
Select tool groups or individual tools to limit what an MCP client loads.

Use Cases

Research assistants can search for current sources and retrieve page content when training data is out of date.
E-commerce and market analysts can collect product details, reviews, company information, or competitor pricing from supported sources.
Social media teams can gather structured profiles, posts, comments, and engagement-related data across supported platforms.
Coding agents can look up current npm and PyPI package metadata, or retrieve files from GitHub repositories.
Agent developers can use remote browser controls when a site requires interaction or dynamic page rendering.

Project Facts

Language: JavaScript
License: MIT
Stars: 2.7k
Forks: 331
Topics: ai-agents, ai-integrations, anti-bot-detection, browser-automation, data-collection, data-extraction, llm, mcp, mcp-server, modelcontextprotocol, scraping, scraping-tools, structured-data, web-crawling, web-data, web-scraping
Archived: No

Getting Started

A local MCP client can launch the server with npx and an API token:

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

A hosted endpoint is also available. See the README for client-specific setup, tool selection, and configuration details.

Alternatives

deepscrape: DeepScrape is self-hosted and combines HTTP fetching, Playwright, and optional LLM extraction, while brightdata-mcp uses Bright Data's managed web-data services.
stagehand: Stagehand is a browser automation SDK that can run locally or on Browserbase, while brightdata-mcp provides MCP tools for search, scraping, extraction, and remote browsing.
Browserable: Browserable is a self-hostable browser automation library for agent-controlled website interactions, while brightdata-mcp also provides web search and structured extraction through Bright Data.

Considerations

An API token and Bright Data account are required, including when using the hosted server.
The README describes a free tier of 5,000 requests per month; further usage is pay-as-you-go, subject to the provider's pricing and account settings.
The MCP server depends on Bright Data's hosted services and infrastructure, rather than providing an independent scraping network.
Structured-data tools expect platform-specific URL formats, and some tool groups must be enabled explicitly.
The repository is MIT-licensed, but access to the underlying data services is governed by Bright Data's terms and pricing.