mcp-popup-ui
Allows GitHub Copilot in VS Code to present interactive popups for user input.
Enables interactive popup-based user input collection when used with Ollama and compatible frontends.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mcp-popup-ui Show me a popup to select my preferred JavaScript framework: React, Vue, or Svelte."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
mcp-popup-ui
An MCP (Model Context Protocol) server that lets AI assistants ask you questions through a visual popup in your browser - instead of just printing text options.

š¢ A Note on the Future
This project started as an experimental solution to a gap in the MCP ecosystem: letting AI assistants collect user input through a visual interface rather than plain text. It works today with any MCP-compatible client.
The MCP specification now includes MCP Apps - an official extension that embeds interactive UIs directly in the chat conversation. This provides a better user experience (no browser popup, no context switching) and will eventually be the recommended approach.
For now, mcp-popup-ui remains useful because MCP Apps support is still limited to a few clients (Claude, VS Code Insiders, Goose, Postman, MCPJam). Once MCP Apps becomes widely available, consider migrating to an MCP Apps-based solution for the best experience.
What This Does
When you chat with an AI assistant (like GitHub Copilot, Claude, or a local LLM), the AI sometimes needs your input - for example, "Which framework do you want?" or "Select the features to include."
Without this tool, the AI would print a numbered list and ask you to type your choice. With mcp-popup-ui, a clean popup opens in your browser where you can click your selection directly.
Two tools are provided:
Tool | Purpose | UI Element |
| Pick exactly one option | Radio buttons |
| Pick one or more options | Checkboxes |
Related MCP server: Cloudflare Playwright MCP
Installation
Install globally via npm:
npm install -g mcp-popup-uiOr run directly without installing:
npx y- mcp-popup-uiRequirements: Node.js 18 or higher.
Setup
Choose the setup guide for your AI application:
VS Code (GitHub Copilot)
Open your project folder in VS Code
Create or edit the file
.vscode/mcp.json:{ "servers": { "popup-ui": { "command": "npx", "args": ["mcp-popup-ui"] } } }Restart VS Code or reload the window
LM Studio
Open LM Studio settings
Navigate to the MCP Servers section
Add a new server with these settings:
{ "mcp-popup-ui": { "command": "npx", "args": ["mcp-popup-ui"] } }Enable the server and start a new chat
Ollama (via Open WebUI or similar)
Ollama itself does not natively support MCP. However, you can use it with frontends that support MCP, such as Open WebUI with MCP plugins. The configuration depends on your specific frontend - consult its documentation for adding MCP servers.
Claude Desktop
Add to your Claude Desktop configuration file:
Windows:
%APPDATA%\Claude\claude_desktop_config.jsonmacOS:
~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"popup-ui": {
"command": "npx",
"args": ["mcp-popup-ui"]
}
}
}Restart Claude Desktop after saving.
Copilot Instructions (Optional)
To ensure the AI uses the popup tools automatically instead of listing options in text, add this to your project's .github/copilot-instructions.md:
## User Input Collection
Use MCP tools for user choices:
- `ask_user` - single selection (radio buttons)
- `ask_user_multiple` - multiple selection (checkboxes)
Use these tools when presenting options like framework choices, implementation approaches, or any list of alternatives.Tool Reference
ask_user
Displays a popup with radio buttons. The user picks exactly one option.
Parameters:
Parameter | Required | Description |
| Yes | Array of options (minimum 2). Each option has a |
| No | Heading displayed above the options |
| No | Additional text displayed below the title |
| No | If |
| No | Label for the custom input option (default: "Other") |
Example call:
{
"options": [
{ "label": "React", "description": "Component-based UI library" },
{ "label": "Vue", "description": "Progressive JavaScript framework" },
{ "label": "Svelte", "description": "Compile-time framework", "recommended": true }
],
"title": "Choose a Frontend Framework",
"description": "Select one framework for your project."
}Response:
{
"action": "submit",
"selection": "Svelte"
}If the user clicks Skip:
{
"action": "skip"
}ask_user_multiple
Displays a popup with checkboxes. The user picks one or more options.
Parameters:
Parameter | Required | Description |
| Yes | Array of options (minimum 2). Each option has a |
| No | Heading displayed above the options |
| No | Additional text displayed below the title |
| No | If |
| No | Label for the custom input option (default: "Other") |
Example call:
{
"options": [
{ "label": "TypeScript", "recommended": true },
{ "label": "ESLint" },
{ "label": "Prettier" },
{ "label": "Jest" }
],
"title": "Select Project Features",
"description": "Choose all features to include."
}Response:
{
"action": "submit",
"selections": ["TypeScript", "ESLint", "Prettier"]
}Additional Features
Skip button: Users can skip any question without selecting an option
Comments field: Users can add additional notes with their selection
Explanation request: Users can ask for more details about an option before deciding
Markdown support: Option descriptions support Markdown formatting
Documentation
Tool Design Best Practices - Research on LLM tool design patterns
Contributing
See CONTRIBUTING.md for development setup and guidelines.
License
MIT
Available Tools
2 toolsask_userA
Ask the user to choose exactly ONE option from a list. Use this tool instead of listing options in your text response whenever you need the user to make a decision.
WHEN TO USE THIS TOOL:
You are about to list numbered options and ask "which do you prefer?"
You need user confirmation on a specific choice before proceeding
The user must pick one mutually exclusive option (e.g., "Which approach should I implement?", "Which file should I modify?", "What framework do you want?")
EXAMPLES OF WHEN TO USE:
"Should I use Option A, B, or C?" ā Use this tool
"Which implementation approach?" ā Use this tool
"What programming language?" ā Use this tool
"Pick a template to scaffold" ā Use this tool
ā ļø CRITICAL: Each parameter MUST be passed SEPARATELY. Do NOT combine them!
ā WRONG (options contains title/description - WILL FAIL): options: [{"label": "A"}, {"label": "B"}], "title": "...", "description": "..."
ā CORRECT (each parameter separate): options: [{"label": "A"}, {"label": "B"}] title: "Pick one" description: "Choose your preference"
PARAMETERS:
options (REQUIRED): Array of objects, each with "label" (required) and optional "description", "recommended"
title (optional): String displayed above options
description (optional): String displayed below title
allow_other (optional): Boolean to allow custom input
The tool opens a popup in the user's browser and waits for their selection. If allow_other is true, users can enter custom text if predefined options don't fit. Returns the selected option as a string, or indicates if the user skipped.
Inputs: options (required), title, allow_other, description, +1 more. Returns: action, comments, selection, explainOption, +1 more
ask_user_multipleA
Ask the user to choose ONE OR MORE options from a list. Use this tool instead of listing options in your text response whenever the user can select multiple items.
WHEN TO USE THIS TOOL:
You are about to list items and ask "which ones do you want?" or "select all that apply"
The user can pick multiple non-exclusive options (e.g., "Which features to include?", "Which files to modify?", "What languages do you know?")
You need to gather multiple preferences at once
EXAMPLES OF WHEN TO USE:
"Which features do you want?" ā Use this tool
"Select the files to include" ā Use this tool
"What integrations should I add?" ā Use this tool
"Pick all applicable tags" ā Use this tool
ā ļø CRITICAL: Each parameter MUST be passed SEPARATELY. Do NOT combine them!
ā WRONG (options contains title/description - WILL FAIL): options: [{"label": "A"}, {"label": "B"}], "title": "...", "description": "..."
ā CORRECT (each parameter separate): options: [{"label": "A"}, {"label": "B"}] title: "Select features" description: "Choose all that apply"
PARAMETERS:
options (REQUIRED): Array of objects, each with "label" (required) and optional "description", "recommended"
title (optional): String displayed above options
description (optional): String displayed below title
min_selections (optional): Minimum selections required
max_selections (optional): Maximum selections allowed
allow_other (optional): Boolean to allow custom input
The tool opens a popup in the user's browser and waits for their selections. If allow_other is true, users can add custom text if predefined options don't fit. Returns an array of selected options, or indicates if the user skipped.
Inputs: options (required), title, allow_other, description, +1 more. Returns: action, comments, selections, explainOption, +1 more
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelogā¦
TDQS
Scored across 2 tools
The two tools are clearly differentiated by singular vs. plural selection. Ask_user restricts to exactly one option, while ask_user_multiple allows multiple selections, with descriptions emphasizing this distinction.
Both tools follow a consistent snake_case verb_noun pattern: ask_user and ask_user_multiple. The naming clearly communicates the shared purpose and the single/multiple distinction.
With only 2 tools, the server feels thin, especially given the broader 'popup-ui' name. The two selection modes cover a narrow niche, and while they are well-defined, the overall surface is minimal.
The tools handle single and multiple choice popups, but other common popup interactions such as free-form text input, simple confirmations, or informational dialogs are missing. Agents must rely on allow_other for custom input, which is not ideal, leaving notable gaps.
Related MCP Connectors
Control a real browser from AI agents: navigate, click, type, screenshot, read the page.
Web search, data extraction and controlled browser automation for AI assistants via OAuth.
Let your AI agent ask a person for a decision, a date or a photo, and get a structured answer back.
Give any AI agent a way to ask a person ā for approval, a decision, an answer or a review.
Related MCP Servers
- FlicenseNot gradedqualityDmaintenanceEnables AI assistants to automate web browsers through Playwright, providing capabilities for navigation, content extraction, form filling, screenshot capture, and JavaScript execution. Supports multiple browser engines with comprehensive error handling and security features.1-
- AlicenseNot gradedqualityNot gradedmaintenanceEnables AI assistants to control a browser for web automation tasks like navigation, typing, clicking, and taking screenshots.Apache 2.0
- FlicenseNot gradedqualityDmaintenanceEnables AI assistants to control a browser for web automation tasks like navigation, typing, clicking, and taking screenshots, with optional API key authentication.-
- AlicenseNot gradedqualityNot gradedmaintenanceEnables AI assistants to control a browser for web automation tasks such as navigation, typing, clicking, and taking screenshots.Apache 2.0