TgeBrowser MCP Server

中文文档

A Model Context Protocol (MCP) server for the TgeBrowser Local API. Enables AI assistants to interact with TgeBrowser — creating, managing, and automating browser environments via the local REST API.

Requirements

• TgeBrowser installed and running
• Node.js 18 or higher

MCP Server Setup

Claude Desktop

Edit the config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%/Claude/claude_desktop_config.json

macOS / Linux:

{
  "mcpServers": {
    "tgebrowser": {
      "command": "npx",
      "args": ["@tgebrowser/mcp"]
    }
  }
}

Windows:

{
  "mcpServers": {
    "tgebrowser": {
      "command": "cmd",
      "args": ["/c", "npx", "@tgebrowser/mcp"]
    }
  }
}

Claude Code

claude mcp add tgebrowser -- npx @tgebrowser/mcp

Cursor

{
  "mcpServers": {
    "tgebrowser": {
      "command": "npx",
      "args": ["@tgebrowser/mcp"]
    }
  }
}

Configuration

Port

Default port is 50326. Override via --port arg or PORT env var.

{
  "mcpServers": {
    "tgebrowser": {
      "command": "npx",
      "args": ["@tgebrowser/mcp", "--port", "50326"]
    }
  }
}

API Key

Provide via --api-key arg or API_KEY env var. Sent as Authorization: Bearer header.

{
  "mcpServers": {
    "tgebrowser": {
      "command": "npx",
      "args": ["@tgebrowser/mcp", "--api-key", "your-api-key"]
    }
  }
}

Combined Example

{
  "mcpServers": {
    "tgebrowser": {
      "command": "npx",
      "args": ["@tgebrowser/mcp", "--port", "50326", "--api-key", "your-api-key"]
    }
  }
}

Skills Support

This project includes intelligent skills for AI assistants to operate TgeBrowser more efficiently.

What are Skills?

Skills provide AI assistants with:

• Smart defaults (auto-fill missing parameters)
• Error prevention (check status before operations)
• Intent inference (understand what user wants)
• Workflow optimization (automate multi-step processes)

Installation

Quick Install:

git clone https://github.com/你的用户名/tgebrowser-mcp-server.git
cp -r tgebrowser-mcp-server/skills/tgebrowser ~/.config/claude/skills/

Detailed Instructions:

• Skills Installation Guide - Multiple installation methods
• OpenClaw Setup Guide - OpenClaw specific configuration

Using Skills

For Claude Code / OpenClaw:

See OPENCLAW_SETUP.md for detailed setup instructions.

Quick setup:

1. Copy or link skills/tgebrowser to your AI assistant's skills directory
2. The AI will automatically use intelligent workflows when operating TgeBrowser

Skills Documentation:

• Main Skill Guide - Complete usage guide
• Environment Management
• Fingerprint Config
• Proxy Config

Available Tools

Browser Management

Tool	Description
create-browser	Create a browser environment
update-browser	Update a browser environment
delete-browser	Delete a browser environment
open-browser	Open a browser environment, returns WebSocket URL for automation
close-browser	Close a browser environment
close-all-profiles	Close all opened browser environments
get-browser-list	List browser environments
get-opened-browser	List currently opened browser environments
get-browser-active	Get active browser environment info
get-profile-cookies	Get cookies of a browser environment
get-profile-ua	Get User-Agent of a browser environment
new-fingerprint	Generate a new random fingerprint
delete-cache	Clear local cache of a browser environment
get-group-list	List browser groups
get-proxy-list	List proxies
check-status	Check TgeBrowser API availability

Browser Automation

After calling open-browser, use connect-browser-with-ws with the returned ws URL to enable automation.

Tool	Description
connect-browser-with-ws	Connect to an opened browser via WebSocket CDP URL
open-new-page	Open a new tab
navigate	Navigate to a URL
screenshot	Take a screenshot of the current page
get-page-visible-text	Get visible text content of the page
get-page-html	Get HTML content of the page
click-element	Click an element by CSS selector
fill-input	Fill an input field
select-option	Select an option from a dropdown
hover-element	Hover over an element
scroll-element	Scroll an element
press-key	Press a keyboard key
drag-element	Drag an element to a target
iframe-click-element	Click an element inside an iframe
evaluate-script	Execute JavaScript in the browser

Once connected, you can talk to your AI assistant naturally. Here are some examples:

Create & manage environments

Create a Windows environment using Chrome 143 kernel, no proxy

Batch create 5 browser environments with random fingerprints on Windows

Rename browser environment ID 12345 to "work-account-01"

List the 10 most recently opened browser environments

Delete all environments whose name contains "test"

Open & automate

Open environment 12345, navigate to https://example.com, and take a screenshot

Open environment 12345, fill in the login form with username admin and password 123456, then click the login button

Open environment 12345 and get all links on the page

Open environments 101, 102, 103 in sequence, navigate each to https://example.com and take a screenshot

Fingerprint & proxy

Generate a new random fingerprint for environment 12345

Check the current User-Agent of environment 12345

Update the proxy of environment 12345 to socks5://127.0.0.1:1080

Groups & proxies

Create a group named "overseas-accounts"

List all proxies and find those of type socks5

Move environments 101, 102, 103 to the "overseas-accounts" group

Development

git clone https://github.com/tgebrowser/tgebrowser.git
cd tgebrowser
pnpm install
pnpm run build

For local development:

{
  "mcpServers": {
    "tgebrowser": {
      "command": "node",
      "args": ["<your-project-path>/tgebrowser/packages/mcp/build/server.js"]
    }
  }
}