User Prompt MCP
About
An MCP server for Cursor that enables requesting user input during generation process.
Details
- License
- MIT
Explore
- User Input Prompting: Allows the AI to ask for more information during generation
- Simple GUI: Presents input prompts in a dialog box with text wrapping
- Cross-Platform: Should work on both Linux (tested) and macOS
- Stdio Transport: Integration with Cursor via stdio
Setting up with Highlight
This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:
- Download and install Highlight from highlightai.com/download
- Navigate to the plugins tab and select "Add Custom Plugin"
-
Configure the plugin with the settings below
Plugin Name
User Prompt MCPCommand (node, npx, python, etc.)Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.
- Enable "Start Automatically" if you want the plugin to start when Highlight launches
From the repository
- For GUI functionality:
- Linux: zenity
- macOS: osascript (built-in)
The easiest way to install user-prompt-mcp is using our installation script, which automatically downloads the appropriate binary for your system:
curl -sSL https://raw.githubusercontent.com/nazar256/user-prompt-mcp/main/install.sh | bash
curl -sSL https://raw.githubusercontent.com/nazar256/user-prompt-mcp/main/install.sh | bash -s -- -v v1.0.0
The installation script will:
- Detect your OS and architecture (Linux or macOS, amd64 or arm64)
- Download the appropriate pre-compiled binary
- Install it to a suitable directory in your PATH
- Make it executable
If you prefer to build from source (requires Go 1.18+):
go install github.com/nazar256/user-prompt-mcp/cmd/user-prompt-mcp@latest
You can also download pre-compiled binaries directly from the Releases page.
1. Open Cursor and go to Settings
2. Navigate to the MCP Servers section and add a new server by editing JSON:
{
"mcpServers": {
"prompt": {
"command": "user-prompt-mcp"
}
}
}
3. Save and refresh
4. Add to global rules something like: "Important - do not finish responding until explicitly told to do so, instead always use MCP input_user_prompt and wait for response in all cases, either you feel stuck, you have question or you finished work on a prompt - always communicate with the user using this MCP."
By default, the server will wait 20 minutes for user input before timing out. You can customize this timeout using:
- Command line flag: --timeout <seconds>
user-prompt-mcp --timeout 600 # Set timeout to 10 minutes
- Environment variable:
USER_PROMPT_TIMEOUT=<seconds> export USER_PROMPT_TIMEOUT=1800 # Set timeout to 30 minutes
user-prompt-mcp
Now when using Cursor, the AI can request additional input from you without ending its generation.
Claude Desktop / Cursor
Paste into your MCP client config file to install this server.
{
"mcpServers": {
"user prompt mcp": {
"ask": {
"command": "~/.local/bin/user-prompt-mcp"
}
}
}
}McpServers
{
"ask": {
"command": "~/.local/bin/user-prompt-mcp"
}
}A Model Context Protocol (MCP) server for Cursor that enables requesting user input during generation. This is mostly AI-generated code.
Overview
This project implements an MCP server that allows Cursor (or any MCP-compatible client) to request additional input from users during model generation without ending the generation process. It serves as a bridge between the AI model and the user, creating a more interactive experience.
Features
- User Input Prompting: Allows the AI to ask for more information during generation
- Simple GUI: Presents input prompts in a dialog box with text wrapping
- Cross-Platform: Should work on both Linux (tested) and macOS
- Stdio Transport: Integration with Cursor via stdio
Installation
Prerequisites
- For GUI functionality:
- Linux: zenity
- macOS: osascript (built-in)
Quick Install (Recommended)
The easiest way to install user-prompt-mcp is using our installation script, which automatically downloads the appropriate binary for your system:
```bash
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
