This project provides a local, cross-platform Model Context Protocol (MCP) server for Node.js/TypeScript, enabling direct user interaction with AI agents. It's designed for developers and power users who need their AI assistants to request input, display notifications, or initiate command-line chats on their local machine, enhancing workflows like interactive configuration or pair programming.

How It Works

The server exposes several MCP tools, including request_user_input for prompts with optional predefined choices, message_complete_notification for OS-level alerts, and start_intensive_chat/ask_intensive_chat/stop_intensive_chat for persistent command-line sessions. This approach allows LLMs to break out of purely text-based, asynchronous communication and engage users directly on their local environment, facilitating more natural and controlled AI-assisted tasks.

Quick Start & Requirements

• Install/Run: Via npx -y interactive-mcp (add to client config).
• Prerequisites: Node.js (version specified in package.json), pnpm for development.
• Configuration: Add to claude_desktop_config.json, mcp.json (Cursor), or VS Code settings.json.
• Docs: Blog Post, Demo Video.

Highlighted Details

• Supports interactive prompts, OS notifications, and command-line chat sessions.
• User input prompts have a configurable timeout (default 30 seconds).
• Tools can be selectively disabled via command-line flags.
• Designed for integration with clients like Claude Desktop and Cursor.

Maintenance & Community

This project is in its early stages. Contribution guidelines are basic. No community links (Discord/Slack) or roadmap are provided.

Licensing & Compatibility

MIT License. Compatible with commercial and closed-source applications.

Limitations & Caveats

The project is noted as being in its early stages. macOS recommendations suggest specific Terminal.app profile settings for managing shell windows, indicating potential complexities in window management.