The web data layer for AI agents

Production-grade access to the open web.
Search, scrape, browse, and extract from any public site - at any scale, in any geo, without getting blocked.

Connect Bright Data to Claude Code, Cursor, or any MCP-compatible client in one line:

claude mcp add --transport http brightdata https://mcp.brightdata.com/mcp?token=YOUR_API_TOKEN

Prefer raw HTTP? Hit the API directly:

curl -X POST https://api.brightdata.com/request \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"zone": "unblocker", "url": "https://example.com", "format": "raw"}'

Or use the SDK:

from brightdata import SyncBrightDataClient

with SyncBrightDataClient(api_token="YOUR_API_TOKEN") as client:
    result = client.scrape_url(url="https://example.com")
    print(result)

Free for agents. MCP and CLI users get 5,000 requests/month - no setup beyond the API key. Grab one →

AI agents are only as good as the data they can reach.

We're the web data infrastructure behind some of the world's most demanding AI systems — billions of requests a day, every public site, every geography, no breakage. While other tools work great until the page actually has a CAPTCHA, geo-block, or anti-bot stack, we handle the parts of the web that break everyone else.

If your agent needs to read the open web in production, this is the layer underneath it.

One MCP for the web. Search, navigate, extract, and unlock any public site from inside Claude, Cursor, Codex, or any MCP client. Used by 150K+ agent calls per week.

Give coding agents real-time web data without leaving the shell. Install once, and Claude Code, Codex, and Gemini CLI can search, scrape, and crawl autonomously.

npx skills add brightdata/cli

The site you're trying to scrape is fighting back? We solve fingerprinting, CAPTCHAs, IP rotation, retries, and anti-bot stacks in real time — single API call, you get the page.

Real user search results from Google, Bing, and every major engine. Returns clean JSON or HTML, with full geo and device targeting. Ranked exactly the way a real person would see them.

Drop-in replacement for Playwright and Puppeteer with proxy rotation, CAPTCHA solving, and anti-detection built in. Render JavaScript-heavy sites at scale, the way an agent needs to.

Skip the scraping infrastructure entirely. Subscribe to fresh structured datasets — products, jobs, profiles, news, reviews — delivered via API, webhook, S3, or your warehouse of choice.

A few patterns we see most often:

• Research agents that browse, read, and synthesize across hundreds of sources
• Price intelligence workflows that monitor SKUs across regions in real time
• Lead enrichment pipelines that pull live company, role, and contact signals
• Training data collection at petabyte scale for foundation models
• Vertical agents for travel, real estate, e-commerce, and finance that need fresh inventory

Browse example apps and integrations →

We're built into the agent stack you're already using:

LangChain · LangGraph · LlamaIndex · CrewAI · Vercel AI SDK · n8n · Make · Zapier · IBM watsonx · AWS Bedrock · Snowflake · Databricks

See all integrations →