browser-control-mcp  by eyalzh

Browser control for AI agents via secure MCP server and extension

Created 1 year ago
258 stars

Top 98.0% on SourcePulse

GitHubView on GitHub
Project Summary

AI agents can gain controlled access to a user's browser through the browser-control-mcp project, which pairs an MCP server with a Firefox extension. This system enables AI assistants to manage tabs, search browsing history, and read webpage content, offering a secure alternative to traditional web automation tools for AI-driven browser interaction.

How It Works

The project employs a two-component architecture: a Firefox browser extension and a local MCP (Message Passing Control) server. The extension interfaces directly with browser functionalities like tab management, history, and content retrieval. It communicates securely with the MCP server using a local-only connection secured by a shared secret. This design prioritizes user privacy and security by avoiding remote data collection and requiring explicit, per-domain user consent for accessing sensitive webpage data or performing actions like text highlighting.

Quick Start & Requirements

Installation can be done via the Firefox Add-ons website or by loading a temporary add-on (manifest.json from the firefox-extension folder). Alternatively, clone the repository and run npm install followed by npm run build to build from source. The MCP server requires configuration within tools like Claude Desktop (claude_desktop_config.json), specifying the server command, arguments, and a secret key obtained from the extension's options page. Docker installation is also supported. Prerequisites include Firefox, Node.js/npm (for building), and potentially Claude Desktop.

Highlighted Details

  • Tab Management: Supports opening, closing, listing, grouping (with name/color), and reordering tabs.
  • History Search: Enables reading and searching the browser's history.
  • Content Access: Allows reading webpage text content and links, subject to user consent per domain.
  • Text Interaction: Facilitates finding and highlighting text within a browser tab, also requiring user consent.
  • Security Focus: Employs local-only connections, shared secrets, consent-based data access, and an extension-side audit log, explicitly avoiding arbitrary scripting and page modification.

Maintenance & Community

No specific details regarding maintainers, community channels (like Discord/Slack), or project roadmap were provided in the README.

Licensing & Compatibility

The license type and any associated restrictions for commercial use or closed-source linking are not explicitly stated in the provided documentation. The system is designed for Firefox and integrates with tools like Claude Desktop.

Limitations & Caveats

This project is explicitly marked as "experimental" and should be used with caution. It does not support direct web page modification, arbitrary scripting, or general page interactions, limiting its scope compared to broader web automation frameworks. User consent is mandatory for accessing webpage content and performing text highlighting actions.

Health Check
Last Commit

3 months ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
12 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.