Report Abuse

Basic Information

interactive-mcp is a local Node.js/TypeScript MCP server that lets large language models running in MCP-capable clients interact directly with a user’s operating system for human-in-the-loop workflows. It is designed to run alongside MCP clients such as Claude Desktop, Cursor, or VS Code so an LLM can request user input, show notifications, and open persistent command-line chat sessions. The server advertises a small set of interactive tools over the Model Context Protocol and is intended for scenarios where clarification, confirmation, or incremental feedback from a human is required during automated tasks. The project is in early stages and targets developers and power users who configure MCP client JSON to launch the server locally via npx.

Links

Categorization

App Details

Features
Exposes explicit MCP tools including request_user_input for prompting users with optional predefined choices, message_complete_notification for OS notifications, and a trio of intensive chat tools to start, ask within, and stop a persistent CLI chat. Provides client configuration examples for Claude Desktop, Cursor, and VS Code showing how to invoke the server with npx and pass command-line flags. Supports command-line options such as --timeout to control prompt timeouts and --disable-tools to selectively disable functionality. Includes development scripts and guidance: pnpm install, pnpm start, pnpm build, pnpm lint, and pnpm format. Documentation contains demos and usage scenarios illustrating interactive prompts, notifications, and intensive chat.
Use Cases
The server prevents agents from guessing by enabling explicit, programmatic user interaction during automated workflows. It is useful for interactive setup and configuration, gathering feedback during code generation or modification, clarifying instructions in pair programming, and confirming actions before making changes. By presenting predefined options and enforcing prompt timeouts, it helps reduce unexpected behavior from LLM clients and encourages confirmation and clarification steps. Local execution gives the server direct access to OS notifications and terminal windows, enabling desktop notifications and persistent command-line chats that improve transparency and control when an LLM needs human input.

Please fill the required fields*