Browser

Browser MCP Server provides browser automation powered by MCP, enabling you to navigate pages, extract content, interact with elements, and perform automated tests in a secure, configurable environment. It leverages Puppeteer to offer reliable web automation within an MCP workflow, making it easy to integrate browser-based tasks into your automation pipelines.

How to use

You can use the Browser MCP Server with an MCP client to perform end-to-end browser automation tasks. Common workflows include navigating to URLs, extracting page sources and metadata, taking screenshots, and converting page content into structured Markdown. You can interact with page elements by clicking, typing, selecting options, and waiting for elements to appear. Test-like actions such as checking element presence, validating content, and running small scripts are also supported. For search needs, you can perform a Baidu search to retrieve the latest articles and information. All actions run inside a sandboxed browser with domain restrictions for safety, and you can customize behavior through configuration options.

How to install

# Prerequisites
# - Node.js installed (npm comes with Node)
# - Git installed

# Clone the repository
git clone https://github.com/yourusername/browser-mcp-server.git
cd browser-mcp-server

# Install dependencies
npm install

# Build the project
npm run build

# Optional: run in development mode
npm run dev

Additional sections

Configuration and runtime behavior are controlled through a configuration file. You can adjust browser settings, security options, and default actions to tailor the server to your environment. The default configuration is defined in the source’s config directory.

Usage with MCP clients involves declaring the server in your MCP configuration as a local (stdio) server. The following example shows the required command and arguments to run the server from your client configuration.

{
  "mcpServers": {
    "browser": {
      "command": "node",
      "args": ["path/to/browser-mcp-server/dist/index.js"]
    }
  }
}

Notes

This server is designed to work in conjunction with an MCP client and is intended to be run as a development or production service within your automation environment. The project follows a standard Node.js build and run flow, including a dev mode for rapid iteration and a test suite to verify functionality.

Available tools

navigate

Navigate to a URL, refresh the page, or move backward/forward in the browser history.

extract_content

Extract page source, metadata, screenshots, and textual content from the current page.

format_md

Convert the current page content into a structured Markdown representation for easy consumption.

interact_click

Click elements matched by selectors or text to trigger UI interactions.

interact_input

Type into inputs, fill forms, and simulate user keystrokes.

interact_select

Choose options from select elements or custom dropdowns.

element_exists

Check whether a specific element exists on the page to drive conditional logic.

verify_content

Validate that certain text or metadata appears on the page to confirm expected results.

execute_script

Run custom scripts in the browser context for advanced interactions.

search_baidu

Perform a Baidu search to fetch the latest articles and information.

security_sandbox

Enforce sandboxed browser execution with domain restrictions to enhance safety.

config_manage

Manage and apply configurable browser options and default behaviors.