Model Context Protocol moderate risk

Interactive

Interactive MCP server enabling LLMs to request user input, send notifications, and initiate command-line chat sessions on the user's local machine.

Connections & Capabilities

Connects To

GitHub

Capabilities

readwriteexec

Quickstart

Install

npm install -g

Config

{
  "mcpServers": {
    "interactive": {
      "command": "npx",
      "args": ["-y", "interactive-mcp"]
    }
  }
}

Exposed MCP Tools (5)

moderate

request_user_input

Asks the user a question and returns their answer, optionally displaying predefined options.

User input can be manipulated, but the tool itself doesn't directly cause harm.

safe

message_complete_notification

Sends a simple OS notification to the user.

Notifications are generally harmless.

moderate

start_intensive_chat

Initiates a persistent command-line chat session with the user.

Opens a command-line interface, potentially allowing for execution of commands.

moderate

ask_intensive_chat

Asks a question within an active intensive chat session.

Relies on user input within a command-line context.

safe

stop_intensive_chat

Closes an active intensive chat session.

Simply closes a command-line session.

Safety Assessment

This server is relatively safe when used with trusted LLMs that adhere to the guiding principles. However, the ability to execute commands on the local machine introduces a moderate risk if the LLM is compromised or poorly configured. Exercise caution when using with untrusted LLMs.

Requires local execution, limiting external exposure.
User interaction is required for most actions.
No inherent network access or data exfiltration capabilities.
Clear guiding principles for interaction are provided.

Executes commands on the user's local machine.
Relies on user input, which could be malicious.
Improperly configured LLMs could lead to unintended actions.
Lack of built-in sandboxing for command execution.