MCP Catalogs
Homehanzi-browse screenshot

hanzi-browse

by hanzili·161·Score 48

MCP server that enables AI agents to browse local websites with site-specific playbooks for complex interactions.

browser-automationweb-scrapingproductivity
48
Forks
29
Open issues
1 mo ago
Last commit
2d ago
Indexed

Overview

Hanzi Browse is a context layer for browsing agents that provides site playbooks to help AI navigate complex websites. It offers two modes: MCP integration for AI agents and API access for building applications. The server exposes tools for browser automation including browser_start, browser_message, browser_status, browser_stop, and browser_screenshot. It supports 24 major websites including X, LinkedIn, Gmail, GitHub, and others with specialized interaction hints rather than brittle scripts.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

you:Automate data extraction from complex websites that have dynamic content
you:Enable AI agents to perform login workflows and form submissions
you:Integrate browser automation into AI coding assistants for testing and deployment
you:What sites are supported?
you:How does it handle anti-bot detection?

When to choose this

Choose Hanzi Browse when you need to automate browser interactions for AI agents on complex websites with dynamic content and anti-bot measures that standard scrapers can't handle.

When NOT to choose this

Don't choose Hanzi Browse if you need to automate simple sites that standard web scrapers can handle, or if you require open-source licensing (the core components are proprietary).

Tools this server exposes

5 tools extracted from the README
  • browser_start

    Run a browser automation task. Blocks until complete.

  • browser_message

    Send a follow-up message to an existing browser session.

  • browser_status

    Check the progress of an ongoing browser automation task.

  • browser_stop

    Stop a currently running browser automation task.

  • browser_screenshot

    Capture the current browser page as an image.

Comparable tools

puppeteerplaywrightbrowserless

Installation

# Install via npx
npx hanzi-browse setup

For Claude Desktop integration, add to Claude Desktop config:

{
  "mcpServers": {
    "hanzi-browse": {
      "command": "npx",
      "args": ["hanzi-browse", "mcp"]
    }
  }
}

FAQ

What sites are supported?
Hanzi Browse currently supports 24 sites including X, LinkedIn, Gmail, GitHub, Notion, Figma, Slack, Reddit, Amazon, eBay, Walmart, Target, Zillow, Apartments.com, Craigslist, Indeed, Google Docs, Sheets, Calendar, Drive, ChatGPT, Claude.ai, and Stack Overflow.
How does it handle anti-bot detection?
Instead of brittle scripts that break when websites change, Hanzi Browse provides hints to the LLM about how to navigate specific sites. The model stays in control and can adapt when the DOM changes, without requiring rebuilding of adapters.

Compare hanzi-browse with

GitHub →

Last updated · Auto-generated from public README + GitHub signals.