Stealth Browser MCP provides AI agents with undetectable, real-browser automation, bypassing anti-bot systems like Cloudflare. It enables AI-driven network analysis and pixel-perfect UI cloning, allowing agents to interact with any website reliably. This tool supercharges AI capabilities for data extraction, reverse engineering, and complex web automation tasks.

How It Works

This project leverages the Chrome DevTools Protocol (CDP) and a modular architecture built on nodriver and FastMCP. Its core innovation lies in AI-generated network hooks and UI cloning capabilities, allowing agents to dynamically adapt to website defenses. The system offers full network debugging via AI chat and supports customizable installations, reducing resource footprint by disabling unused modules.

Quick Start & Requirements

• Install: Clone repo, create/activate Python virtual environment, pip install -r requirements.txt, then integrate with an MCP-compatible AI agent (e.g., Claude) via CLI or manual configuration.
• Prerequisites: Python 3.x, Git, a compatible browser (Chrome, Chromium, Edge).
• Links: Quickstart, Discord, Roadmap.

Highlighted Details

• Undetectable Automation: Claims high success rates (98.7%) against Cloudflare, CAPTCHAs, and social media blocks.
• AI Integration: Network hooks and UI cloning are AI-driven via chat commands.
• CDP-Accurate Cloning: Pixel-perfect UI element extraction and recreation.
• Advanced Text Input: paste_text() offers 10x faster input than typing.
• Robust Environment Handling: Auto-detects and configures for root, containers, and various platforms.
• Modular Design: 11 tool sections (22-89 tools total), allowing minimal or custom installations.

Maintenance & Community

Actively developed with recent enhancements (v0.2.4). Community support available via Discord. Roadmap is publicly available.

Licensing & Compatibility

• License: MIT.
• Compatibility: Permissive license allows commercial use and integration into closed-source projects.

Limitations & Caveats

Effectiveness of AI features depends on the integrated AI agent's capabilities. Some integration aspects (e.g., FastMCP CLI) are noted as untested by the creator. While robust, complex or unusual environment setups might require manual troubleshooting.