MCP-Browse

What is MCP-Browse?

MCP-Browse is a browser control protocol for automating interactions with web pages, particularly e-commerce sites. It provides a clean interface for navigation, clicking, form input, downloading files, and executing JavaScript, using Protocol Buffers for efficient typed communication.

How to use MCP-Browse?

Install Go 1.18+ and Protocol Buffers compiler, clone the repository, run go mod download, generate Protocol Buffer code with buf generate, then create a client using NewMCPBrowserServiceClient and invoke gesture operations like Navigate, Click, and Input which return streamed responses.

Key features of MCP-Browse

• Gesture-based protocol for browser interactions
• Supports Navigation, Click, Input, Download, ExecuteScript
• Streamed responses for navigation and input gestures
• Uses Protocol Buffers for efficient typed communication
• Focused on automating e-commerce site workflows

Use cases of MCP-Browse

• Automate product page navigation and clicks
• Fill and submit e-commerce forms (add to cart, checkout)
• Download images or files from web pages
• Execute custom JavaScript on pages to extract data
• Handle redirects during page navigation

FAQ from MCP-Browse

What runtime does MCP-Browse require?

Go 1.18+ and the Protocol Buffers compiler are required.

How does MCP-Browse handle page navigation?

It returns a stream with optional redirect events, page content or error info, and a completion event.

What types of browser interactions are supported?

Navigate, Click (by ID or CSS selector), Input (text, checkboxes, dropdowns), Download, and ExecuteScript.

Is MCP-Browse a standalone server or a library?

The README describes a client library; the protocol is server-based but the server implementation is not detailed in the provided text.

How are responses returned for different operations?

Browser gesture operations return streamed responses; utility operations (Download, ExecuteScript) return a single response.