Browser Control MCP Server

@adityasasidhar

リポジトリ

基本情報

カテゴリ

ブラウザ自動化

トランスポート

stdio

作者

adityasasidhar

投稿者

Aditya Sasidhar

Browser Control MCP Server について

Stop feeding your agent raw HTML. BrowserControl gives LLMs true visual grounding with Set of Marks (SoM), letting them click exactly what they see without hallucinating selectors. With built-in developer tools, session recording, and persistent logins, it transforms web automati

#browser-control#browser-automation #developer-tools

Browser Control MCP Server の設定

以下の設定を使って、このサーバーを MCP 対応クライアントに追加してください。

{
  "mcpServers": {
    "browsercontrol": {
      "command": "python",
      "args": [
        "-m",
        "browsercontrol"
      ]
    }
  }
}

Browser Control MCP Server のツール

ツール一覧はまだありません

この掲載には表示できるツール一覧がまだありません。サーバーにツールがないという意味ではありません。対応機能はプロジェクトのドキュメントをご確認ください。

ツールは README から自動的に抽出されます。メンテナーは ## Tools という見出しの下に記載することで、このタブに反映できます。

概要

What is Browser Control?

Browser Control is a production-ready MCP server that gives AI agents real browser capabilities using vision-first interaction. It controls Chromium through Playwright, providing screenshots with numbered interactive elements, persistent sessions, and built-in developer tooling—all locally without external vision APIs.

How to use Browser Control?

Install Python 3.11+ and Playwright (Chromium installed automatically). Add Browser Control as an MCP server in your AI agent’s client. Agents then call tools like click(5) or type_text(3, "hello") to interact with web pages.

Key features of Browser Control

Vision-first interaction via annotated page screenshots
Set of Marks (SoM) with numbered interactive elements
Persistent browser sessions (cookies, login, history)
Built-in developer tools (console, network, JS execution)
Session recording with Playwright traces
Zero extra AI cost (no vision models or selector inference)

Use cases of Browser Control

Autonomous web research and data extraction
Browser-based agent workflows requiring login persistence
Automated form filling and end-to-end testing
Debugging web apps via console logs and network requests

よくある質問

How does Browser Control differ from DOM-based tools?

It uses vision-first interaction based on page screenshots, avoiding fragile DOM queries. This reduces tokens per action by 50–100× and eliminates brittle selector inference.

What are the runtime dependencies?

Python 3.11+, FastMCP, and Playwright (auto-installs Chromium). All execution is local—no external AI or API calls.

Where does browser state and session data live?

All data, including cookies, localStorage, and recording traces, is stored locally on the machine running the MCP server.

What transport does Browser Control use?

It uses the MCP (Model Context Protocol) transport, designed for integration with AI agent clients. No authentication details are specified in the documentation.

Are there any known limits or stability notes?

Browser Control is stable, actively developed, and MIT licensed. No specific limits are documented beyond requiring a local Chromium installation.

「ブラウザ自動化」の他のコンテンツ

Screenshot Scout

screenshotscout

Capture screenshots of webpages as images or PDFs with Screenshot Scout.

Scrapling Fetch MCP

cyberchitta

Helps AI assistants access text content from bot-protected websites. MCP server that fetches HTML/markdown from sites with anti-automation measures using Scrapling.

101

Fastcrw

Fast, lightweight Firecrawl/Tavily alternative in Rust. Web scraper, crawler & search API with MCP server for AI agents. Drop-in Firecrawl-compatible API (/scrape, /crawl, /search). 2.3x faster than Tavily, 1.5x faster than Firecrawl in 1K-URL benchmarks. 6 MB RAM, single binary.

286

Parse

Parse.bot

Parse allows your agent to access thousands of APIs instantly. You can use Parse to build an API for any action on any website and edit any previously existing API in our catalogue.

HTML/CSS to Image

An MCP server for generating images from HTML & CSS or screenshots of URLs using htmlcsstoimage.com.