Brightdata MCP

What luminati-io/brightdata-mcp MCP server does

The luminati-io/brightdata-mcp MCP server gives MCP-compatible agents access to live public web data. Its documented tool set covers web search, page retrieval, structured extraction, browser control, LLM response collection, and package registry information.

Search tools return structured results from Google, Bing, and Yandex. Scraping tools can return a page as Markdown or HTML. Prebuilt extractors provide JSON for services and sites including Amazon, LinkedIn, Instagram, TikTok, YouTube, X, Reddit, Facebook, Crunchbase, and Zillow. The server also includes tools for npm and PyPI metadata such as versions, READMEs, dependencies, and package information.

Browser automation tools support navigation, clicking, typing, screenshots, and reading pages in a remote browser session. Other tools can send prompts to ChatGPT, Grok, and Perplexity and collect their responses as structured data.

How it works

The luminati-io/brightdata-mcp MCP server can be used through Bright Data's hosted endpoint or run locally with the npm package. In either mode, an MCP client invokes the available tools and Bright Data handles the underlying web access. The README states that requests pass through infrastructure for proxy rotation, bot-detection bypass, CAPTCHA solving, rate-limit handling, and geo-targeting.

The hosted endpoint accepts an API token in the URL. Its optional parameters can restrict the connection to selected tool groups or named tools. Local deployments receive the token through the API_TOKEN environment variable.

Setup and configuration

For a hosted deployment, configure an MCP client with:

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

The token comes from Bright Data account settings. The README documents hosted configurations for Claude Desktop, Cursor, Windsurf, and several other clients. A local deployment uses the @brightdata/mcp npm package through npx and sets API_TOKEN in the client configuration.

You can reduce the exposed surface with URL parameters such as groups for tool groups or tools for specific tool names. Bright Data accounts include a recurring free allowance of 5,000 requests per month; usage beyond that is pay as you go according to the service pricing.

Tools and capabilities

The listed MCP groups and tools include search_engine, search_engine_batch, scrape_as_markdown, scrape_batch, discover, and custom. The README describes a broader catalog of 69 tools, including domain-specific extractors and browser operations. discover is intended to find relevant sources for a research question, while batch tools support multiple searches or page requests.

This makes the luminati-io/brightdata-mcp MCP server suitable for current-information research, product and price comparisons, competitor analysis, source collection, package lookup, and agents that need to read protected public pages.

Limitations and notes

The service requires a Bright Data API token, and its request allowance and paid usage are controlled by the Bright Data account. The free allocation renews monthly and unused requests do not roll over. Requests stop when the free allowance is exhausted unless the account has deposited funds for additional usage.

The documented access is for public web data. Results and availability can depend on the target site, the selected tool, and Bright Data's service limits. The hosted option avoids local installation; the local option still requires the API token and an MCP-compatible client.