Overview

The Browserless MCP Server gives AI assistants direct control over headless browsers running in the cloud or self-hosted via Docker. Unlike Puppeteer or Playwright MCP servers that run locally, Browserless handles Chrome in managed infrastructure with built-in stealth detection, captcha solving, and smart scraping — all exposed through a standardised MCP interface.

Built-in MCP:

MCP support is built directly into the Browserless server infrastructure. No separate package — connect via remote MCP endpoint or Docker with MCP transport enabled. Maintained by Browserless.

Key Features

🕷️

Smart Scraping

Extract clean Markdown, HTML, or structured data from any webpage with automatic JS rendering

🛡️

Anti-Detection

Built-in stealth detection bypass, captcha solving, and residential proxy rotation

📄

PDF & Screenshot REST APIs

Generate PDFs, screenshots, and content extractions via simple REST endpoints

🐳

Self-Hosted or Cloud

Run in Docker locally or use the managed cloud service with auto-scaling

🎭

Puppeteer + Playwright

Drop-in replacement — connect your existing Puppeteer or Playwright scripts via WebSocket

🔍

REST APIs

Pre-built endpoints for /content, /pdf, /screenshot, /scrape, /crawl, and /search

Available Tools

Quick Reference

Tool	Purpose	Category
`browserless_smartscraper`	Extract clean content from any URL	Extraction
`browserless_crawl`	Multi-page crawl with depth control	Crawling
`browserless_search`	Web search with optional scraping of results	Search

Detailed Usage

browserless_smartscraper▶

Extract clean, structured content from any URL. Handles JavaScript rendering, anti-bot bypass, and returns Markdown, HTML, or structured data.

{
  "url": "https://example.com",
  "format": "markdown",
  "waitFor": 2000,
  "stealth": true
}

browserless_crawl▶

Crawl multiple pages starting from a root URL with configurable depth and page limits.

{
  "url": "https://docs.browserless.io",
  "maxDepth": 2,
  "maxPages": 50,
  "format": "markdown"
}

browserless_search▶

Search the web and optionally scrape each result page into structured formats.

{
  "query": "latest AI agent frameworks 2026",
  "scrapeResults": true,
  "format": "markdown"
}

Installation

Pull and run the Docker image with MCP transport enabled:

docker run -p 3000:3000 ghcr.io/browserless/chromium

{
  "mcpServers": {
    "browserless": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "http://localhost:3000/mcp"
      ]
    }
  }
}

Direct Docker Run

Run Browserless with MCP transport in a single command:

docker run -p 3000:3000 \
  -e TOKEN=your-token \
  -e CONCURRENT=10 \
  -e TIMEOUT=30000 \
  ghcr.io/browserless/chromium

The MCP endpoint will be available at http://localhost:3000/mcp.

API Token Required for Remote:

Get a free API token by signing up at browserless.io. The open-source Docker image is free for non-commercial use.

Comparison: Browserless vs Browserbase

Feature	Browserless	Browserbase
Deployment	Self-hosted (Docker) + Cloud	Cloud-only
Stealth	Captcha solving, residential proxies	Verified Identity, fingerprinting
AI Actions	Smart scraper, crawl, search	Stagehand AI (natural language)
Session Persistence	Premium feature	Keep Alive mode built-in
REST APIs	PDF, screenshot, content, crawl, search	N/A (Stagehand-focused)
Licensing	SSPL + Commercial (free for non-commercial)	Proprietary (API-based)
MCP Integration	Built-in transport	Separate `@browserbasehq/mcp` package

Security

MCP endpoint secured via token authentication
Browsers run in isolated Docker containers with configurable resource limits
Configurable session timeouts and concurrency caps prevent resource exhaustion
All browser traffic proxied through configurable endpoints
Self-hosted option keeps all data on your infrastructure — no third-party access

Browserless MCP Server